Skip to main content

Head-to-head comparison

Claude Opus 4.6 vs. DeepSeek-V4-Flash

Compare published benchmark results from matching versions and cohorts, API pricing, and technical specifications. Any documented test setup differences remain visible.

shared result series
66
distinct benchmarks
2
latest retrieval
2026-10-15

Change model selection

Specifications

General specifications

Release date, availability, context window, architecture, and published parameter counts at a glance. Undisclosed details remain marked as unavailable.

Specifications for Claude Opus 4.6 and DeepSeek-V4-Flash
AttributeClaude Opus 4.6DeepSeek-V4-Flash
ProviderAnthropicDeepSeek
ReleasedFeb 2026Apr 24, 2026
AvailabilityActiveLimited access
Model typeProprietaryOpen Weights
ParametersNot published284B, 13B active
ArchitectureNot publishedMixture of Experts
Context window1,000,000 tokens1,000,000 tokens
Knowledge cutoffMay 2025Not published
Technical sourcesModel, ContextModel

Pricing

API pricing compared

The table compares official standard API rates per 1M tokens, including published cache pricing. Batch, fast, priority, regional surcharges, tool calls, and cloud platform pricing are excluded.

API pricing for Claude Opus 4.6 and DeepSeek-V4-Flash
Price typeClaude Opus 4.6DeepSeek-V4-Flash
API input$5$0.14
API output$25$0.28
Cache read$0.5Not listed
Cache write (5 min.)$6.25Not listed
Cache write (1 hr.)$10Not listed
Price verified09/22/2026 Source10/17/2026 Source
Claude Opus 4.6DeepSeek-V4-Flash
API inputUSD per 1M tokens, base tier
Claude Opus 4.6$5
DeepSeek-V4-FlashLower price$0.14
API outputUSD per 1M tokens, base tier
Claude Opus 4.6$25
DeepSeek-V4-FlashLower price$0.28

Key to the marks:better valuebetter, but not conclusivetie

Pricing by provider

Direct developer pricing and provider endpoint tariffs routed through OpenRouter are listed separately. Prices are in USD per 1M tokens. Context-dependent direct tiers plus regional, Flex, Priority, and other OpenRouter tariffs remain individually identifiable.

Provider pricing for Claude Opus 4.6 and DeepSeek-V4-Flash
ModelProviderPurchase route and tariffInput per 1M tokensOutput per 1M tokensSource
Claude Opus 4.6AnthropicDirect from developerStandard APILowest listed price: $5Lowest listed price: $25Source09/22/2026
Amazon BedrockVia OpenRouteramazon-bedrockLowest listed price: $5Lowest listed price: $25Source10/17/2026
AnthropicVia OpenRouteranthropicLowest listed price: $5Lowest listed price: $25Source10/17/2026
AzureVia OpenRouterazure/globalLowest listed price: $5Lowest listed price: $25Source10/17/2026
Claude Platform on AWSVia OpenRouterclaude-on-awsLowest listed price: $5Lowest listed price: $25Source10/17/2026
GoogleVia OpenRoutergoogle-vertex/europe$5.5$27.5Source10/17/2026
GoogleVia OpenRoutergoogle-vertex/globalLowest listed price: $5Lowest listed price: $25Source10/17/2026
DeepSeek-V4-FlashAlibabaVia OpenRouteralibaba$0.352$1.056Source10/17/2026
AlibabaVia OpenRouteralibaba/us-east-1$0.374$1.122Source10/17/2026
AtlasCloudVia OpenRouteratlas-cloud/fp4$0.44$1.32Source10/17/2026
BaiduVia OpenRouterbaidu/fp8$0.0598$0.1795Source10/17/2026
BaseTenVia OpenRouterbaseten/fp32$0.13$0.26Source10/17/2026
BaseTenVia OpenRouterbaseten/fp8$0.13$0.26Source10/17/2026
CloudflareVia OpenRoutercloudflare$0.44$1.32Source10/17/2026
CohereVia OpenRoutercohere$0.14$0.28Source10/17/2026
CoreWeaveVia OpenRoutercoreweave/fp8$0.13$0.28Source10/17/2026
DeepInfraVia OpenRouterdeepinfra/fp8$0.06$0.18Source10/17/2026
DigitalOceanVia OpenRouterdigitalocean$0.119$0.238Source10/17/2026
GMICloudVia OpenRoutergmicloud/fp8$0.286$0.858Source10/17/2026
InceptronVia OpenRouterinceptron/fp4$0.13$0.4782Source10/17/2026
Mancer 2Via OpenRoutermancer/fp8$0.2$0.6Source10/17/2026
NovitaVia OpenRouternovita/fp8$0.4092$1.2276Source10/17/2026
OpenInferenceVia OpenRouteropen-inference/fp4$0.0052Lowest listed price: $0.1436Source10/17/2026
ParasailVia OpenRouterparasail/fp8$0.14$0.28Source10/17/2026
PhalaVia OpenRouterphala$0.44$1.32Source10/17/2026
RekaVia OpenRouterreka$0.021$0.528Source10/17/2026
RelaceVia OpenRouterrelace/fp4Lowest listed price: $0.0049$1.28Source10/17/2026
Sail ResearchVia OpenRoutersail-research/fp4$0.1$0.3Source10/17/2026
Sail ResearchVia OpenRoutersail-research/us$0.1$0.3Source10/17/2026
SiliconFlowVia OpenRoutersiliconflow/fp8$0.22$0.66Source10/17/2026
StreamLakeVia OpenRouterstreamlake/fp8$0.088$0.264Source10/17/2026
TogetherVia OpenRoutertogether$0.14$0.28Source10/17/2026
VeniceVia OpenRoutervenice$0.13$0.26Source10/17/2026
WaferVia OpenRouterwafer/fast$0.13$0.23Source10/17/2026

Performance

Shared benchmarks

Only published results with the same benchmark version, task, metric, and comparison cohort are paired. Different reasoning levels, agents, harnesses, or output limits appear directly in the table.

Capability profile from matched benchmarks

Each axis averages directly matched benchmark families. A value of 100 means the stronger value within this pair, not a universal quality score.

255075100CodingReasoningMathematicsKnowledgeFinanceLegalHealthGeneral
Claude Opus 4.6
DeepSeek-V4-Flash
Coding: 100 / 93.9 (2 benchmark families)
Reasoning: 100 / 95.3 (1 benchmark family)
Mathematics: 100 / 94.8 (1 benchmark family)
Knowledge: 100 / 95.3 (1 benchmark family)
Finance: 100 / 95.8 (1 benchmark family)
Legal: 100 / 96.3 (1 benchmark family)
Health: 100 / 96.4 (1 benchmark family)
General: 100 / 95.4 (1 benchmark family)

62 matched rows from 2 benchmark families were considered. The radar shows 8 of 8 comparable categories.

Claude Opus 4.6DeepSeek-V4-Flash
LMArena Code Arena WebDevOverall. Higher is better. Test setup not documented as identical
Claude Opus 4.6Higher measured value1,537.258
DeepSeek-V4-Flash1,430.632
LMArena Text, Style ControlOverall. Higher is better. No documented setup difference
Claude Opus 4.6Winner1,497.663
DeepSeek-V4-Flash1,435.976
LMArena Code Arena WebDevOverall. Higher is better. No documented setup difference
Claude Opus 4.6Winner1,546.362
DeepSeek-V4-Flash1,430.632
LMArena Text, Style ControlOverall. Higher is better. No documented setup difference
Claude Opus 4.6Winner1,503.106
DeepSeek-V4-Flash1,438.654
LMArena Code Arena WebDevWebdev. Higher is better. Test setup not documented as identical
Claude Opus 4.6Higher measured value1,537.258
DeepSeek-V4-Flash1,430.632
LMArena Text, Style ControlBusiness, management and finance. Higher is better. No documented setup difference
Claude Opus 4.6Winner1,499.695
DeepSeek-V4-Flash1,438.829
LMArena Code Arena WebDevWebdev. Higher is better. No documented setup difference
Claude Opus 4.6Winner1,546.362
DeepSeek-V4-Flash1,430.632
LMArena Text, Style ControlBusiness, management and finance. Higher is better. No documented setup difference
Claude Opus 4.6Winner1,500.43
DeepSeek-V4-Flash1,434.084
LMArena Code Arena WebDevWebdev-html. Higher is better. Test setup not documented as identical
Claude Opus 4.6Higher measured value1,546.218
DeepSeek-V4-Flash1,416.219
LMArena Text, Style ControlChinese. Higher is better. No documented setup difference
Claude Opus 4.6Winner1,545.153
DeepSeek-V4-Flash1,470.018
LMArena Code Arena WebDevWebdev-html. Higher is better. No documented setup difference
Claude Opus 4.6Winner1,558.715
DeepSeek-V4-Flash1,416.219
LMArena Text, Style ControlChinese. Higher is better. No documented setup difference
Claude Opus 4.6Winner1,550.027
DeepSeek-V4-Flash1,480.516

Key to the marks:better valuebetter, but not conclusivetie

Benchmark scores for Claude Opus 4.6 and DeepSeek-V4-Flash
Benchmark and taskClaude Opus 4.6DeepSeek-V4-FlashSource
LMArena Code Arena WebDevOverall
Test details
Test setup not documented as identicalNo direct winnerVersion: webdevMetric: WebDev Arena Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6, Harness: Arena Code Arena WebDev
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash-high-preview, Reasoning: high, Harness: Arena Code Arena WebDevClaude Opus 4.6: Arena rating from 20,698 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 10,152 published pairwise votes.
Higher measured value1,537.258
95% confidence interval: ±5.345The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,430.632
95% confidence interval: ±6.942The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlOverall
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 85,445 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 52,016 published pairwise votes.
Winner1,497.663
95% confidence interval: ±3.27The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,435.976
95% confidence interval: ±4.036The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Code Arena WebDevOverall
Test details
No documented difference in the test setupVersion: webdevMetric: WebDev Arena Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6-high, Reasoning: high, Harness: Arena Code Arena WebDev
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash-high-preview, Reasoning: high, Harness: Arena Code Arena WebDevClaude Opus 4.6: Arena rating from 19,443 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 10,152 published pairwise votes.
Winner1,546.362
95% confidence interval: ±5.46The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,430.632
95% confidence interval: ±6.942The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlOverall
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6-high, Reasoning: high, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash-high-preview, Reasoning: high, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 80,827 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 51,814 published pairwise votes.
Winner1,503.106
95% confidence interval: ±3.353The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,438.654
95% confidence interval: ±4.106The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Code Arena WebDevWebdev
Test details
Test setup not documented as identicalNo direct winnerVersion: webdevMetric: WebDev Arena Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6, Harness: Arena Code Arena WebDev
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash-high-preview, Reasoning: high, Harness: Arena Code Arena WebDevClaude Opus 4.6: Arena rating from 20,698 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 10,152 published pairwise votes.
Higher measured value1,537.258
95% confidence interval: ±5.345The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,430.632
95% confidence interval: ±6.942The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlBusiness, management and finance
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 16,859 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 10,448 published pairwise votes.
Winner1,499.695
95% confidence interval: ±5.707The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,438.829
95% confidence interval: ±6.961The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Code Arena WebDevWebdev
Test details
No documented difference in the test setupVersion: webdevMetric: WebDev Arena Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6-high, Reasoning: high, Harness: Arena Code Arena WebDev
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash-high-preview, Reasoning: high, Harness: Arena Code Arena WebDevClaude Opus 4.6: Arena rating from 19,443 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 10,152 published pairwise votes.
Winner1,546.362
95% confidence interval: ±5.46The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,430.632
95% confidence interval: ±6.942The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlBusiness, management and finance
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6-high, Reasoning: high, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash-high-preview, Reasoning: high, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 15,865 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 10,643 published pairwise votes.
Winner1,500.43
95% confidence interval: ±5.84The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,434.084
95% confidence interval: ±6.963The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Code Arena WebDevWebdev-html
Test details
Test setup not documented as identicalNo direct winnerVersion: webdevMetric: WebDev Arena Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6, Harness: Arena Code Arena WebDev
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash-high-preview, Reasoning: high, Harness: Arena Code Arena WebDevClaude Opus 4.6: Arena rating from 2,623 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 1,165 published pairwise votes.
Higher measured value1,546.218
95% confidence interval: ±12.521The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,416.219
95% confidence interval: ±18.033The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlChinese
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 5,821 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 3,005 published pairwise votes.
Winner1,545.153
95% confidence interval: ±8.666The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,470.018
95% confidence interval: ±11.382The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Code Arena WebDevWebdev-html
Test details
No documented difference in the test setupVersion: webdevMetric: WebDev Arena Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6-high, Reasoning: high, Harness: Arena Code Arena WebDev
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash-high-preview, Reasoning: high, Harness: Arena Code Arena WebDevClaude Opus 4.6: Arena rating from 2,474 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 1,165 published pairwise votes.
Winner1,558.715
95% confidence interval: ±13.217The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,416.219
95% confidence interval: ±18.033The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlChinese
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6-high, Reasoning: high, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash-high-preview, Reasoning: high, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 5,591 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 2,975 published pairwise votes.
Winner1,550.027
95% confidence interval: ±8.89The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,480.516
95% confidence interval: ±11.412The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Code Arena WebDevWebdev-react
Test details
Test setup not documented as identicalNo direct winnerVersion: webdevMetric: WebDev Arena Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6, Harness: Arena Code Arena WebDev
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash-high-preview, Reasoning: high, Harness: Arena Code Arena WebDevClaude Opus 4.6: Arena rating from 16,319 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 8,109 published pairwise votes.
Higher measured value1,528.416
95% confidence interval: ±5.908The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,421.593
95% confidence interval: ±7.665The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlCoding
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 23,695 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 15,290 published pairwise votes.
Winner1,547.708
95% confidence interval: ±5.22The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,484.301
95% confidence interval: ±6.214The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Code Arena WebDevWebdev-react
Test details
No documented difference in the test setupVersion: webdevMetric: WebDev Arena Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6-high, Reasoning: high, Harness: Arena Code Arena WebDev
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash-high-preview, Reasoning: high, Harness: Arena Code Arena WebDevClaude Opus 4.6: Arena rating from 15,177 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 8,109 published pairwise votes.
Winner1,538.872
95% confidence interval: ±6.058The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,421.593
95% confidence interval: ±7.665The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlCoding
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6-high, Reasoning: high, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash-high-preview, Reasoning: high, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 20,969 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 15,245 published pairwise votes.
Winner1,548.395
95% confidence interval: ±5.448The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,479.463
95% confidence interval: ±6.25The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlCreative writing
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 14,986 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 8,595 published pairwise votes.
Winner1,478.397
95% confidence interval: ±6.086The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,408.246
95% confidence interval: ±7.512The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlCreative writing
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6-high, Reasoning: high, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash-high-preview, Reasoning: high, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 14,826 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 8,668 published pairwise votes.
Winner1,499.767
95% confidence interval: ±6.246The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,410.845
95% confidence interval: ±7.548The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlEnglish
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 38,162 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 23,136 published pairwise votes.
Winner1,506.394
95% confidence interval: ±4.352The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,445.189
95% confidence interval: ±5.273The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlEnglish
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6-high, Reasoning: high, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash-high-preview, Reasoning: high, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 35,612 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 23,022 published pairwise votes.
Winner1,513.707
95% confidence interval: ±4.48The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,449.707
95% confidence interval: ±5.31The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlEntertainment, sports and media
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 19,048 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 11,221 published pairwise votes.
Winner1,478.55
95% confidence interval: ±5.586The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,404.55
95% confidence interval: ±6.918The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlEntertainment, sports and media
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6-high, Reasoning: high, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash-high-preview, Reasoning: high, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 18,282 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 11,460 published pairwise votes.
Winner1,491.123
95% confidence interval: ±5.751The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,407.781
95% confidence interval: ±6.873The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlExcluding ties
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 64,924 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 39,351 published pairwise votes.
Winner1,516.544
95% confidence interval: ±4.425The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,431.858
95% confidence interval: ±5.294The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlExcluding ties
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6-high, Reasoning: high, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash-high-preview, Reasoning: high, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 61,691 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 39,449 published pairwise votes.
Winner1,522.764
95% confidence interval: ±4.502The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,436.129
95% confidence interval: ±5.336The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlExpert prompts
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 8,806 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 5,416 published pairwise votes.
Winner1,533.95
95% confidence interval: ±7.394The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,457.085
95% confidence interval: ±8.727The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlExpert prompts
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6-high, Reasoning: high, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash-high-preview, Reasoning: high, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 7,356 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 5,338 published pairwise votes.
Winner1,545.814
95% confidence interval: ±7.928The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,466.764
95% confidence interval: ±8.816The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlFrench
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 2,808 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 2,004 published pairwise votes.
Winner1,496.997
95% confidence interval: ±13.146The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,449.161
95% confidence interval: ±14.911The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlFrench
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6-high, Reasoning: high, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash-high-preview, Reasoning: high, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 2,619 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 1,957 published pairwise votes.
Winner1,511.426
95% confidence interval: ±13.557The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,458.886
95% confidence interval: ±15.182The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlGerman
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 1,328 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 841 published pairwise votes.
Winner1,490.887
95% confidence interval: ±17.278The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,425.17
95% confidence interval: ±21.221The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlGerman
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6-high, Reasoning: high, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash-high-preview, Reasoning: high, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 1,284 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 904 published pairwise votes.
Winner1,507.167
95% confidence interval: ±17.309The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,433.291
95% confidence interval: ±20.525The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlHard prompts
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 55,452 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 34,789 published pairwise votes.
Winner1,527.527
95% confidence interval: ±3.977The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,459.087
95% confidence interval: ±4.755The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlHard prompts
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6-high, Reasoning: high, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash-high-preview, Reasoning: high, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 51,483 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 34,719 published pairwise votes.
Winner1,531.566
95% confidence interval: ±4.093The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,458.022
95% confidence interval: ±4.823The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlHard prompts, English
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 25,280 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 16,226 published pairwise votes.
Winner1,529.961
95% confidence interval: ±5.118The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,463.782
95% confidence interval: ±6.043The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlHard prompts, English
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6-high, Reasoning: high, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash-high-preview, Reasoning: high, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 23,143 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 16,101 published pairwise votes.
Winner1,535.921
95% confidence interval: ±5.299The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,464.326
95% confidence interval: ±6.118The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlInstruction following
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 28,622 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 17,984 published pairwise votes.
Winner1,499.444
95% confidence interval: ±4.848The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,428.161
95% confidence interval: ±5.795The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlInstruction following
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6-high, Reasoning: high, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash-high-preview, Reasoning: high, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 26,036 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 18,097 published pairwise votes.
Winner1,512.245
95% confidence interval: ±5.064The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,433.654
95% confidence interval: ±5.864The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlJapanese
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 941 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 653 published pairwise votes.
Winner1,483.858
95% confidence interval: ±21.248The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,398.855
95% confidence interval: ±24.754The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlJapanese
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6-high, Reasoning: high, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash-high-preview, Reasoning: high, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 922 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 640 published pairwise votes.
Winner1,498.901
95% confidence interval: ±21.325The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,426.069
95% confidence interval: ±24.977The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlKorean
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 1,536 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 900 published pairwise votes.
Winner1,465.035
95% confidence interval: ±16.873The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,391.938
95% confidence interval: ±20.897The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlKorean
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6-high, Reasoning: high, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash-high-preview, Reasoning: high, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 1,361 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 949 published pairwise votes.
Winner1,445.732
95% confidence interval: ±17.684The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,376.332
95% confidence interval: ±20.813The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlLegal and government
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 6,844 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 4,152 published pairwise votes.
Winner1,506.526
95% confidence interval: ±8.092The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,448.281
95% confidence interval: ±9.836The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlLegal and government
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6-high, Reasoning: high, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash-high-preview, Reasoning: high, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 6,593 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 4,099 published pairwise votes.
Winner1,510.822
95% confidence interval: ±8.22The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,457.757
95% confidence interval: ±9.897The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlLife, physical and social science
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 14,216 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 8,416 published pairwise votes.
Winner1,518.741
95% confidence interval: ±6.004The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,456.243
95% confidence interval: ±7.368The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlLife, physical and social science
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6-high, Reasoning: high, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash-high-preview, Reasoning: high, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 13,449 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 8,489 published pairwise votes.
Winner1,528.541
95% confidence interval: ±6.139The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,459.964
95% confidence interval: ±7.4The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlLonger queries
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 37,162 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 23,382 published pairwise votes.
Winner1,516.98
95% confidence interval: ±4.682The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,447.798
95% confidence interval: ±5.56The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlLonger queries
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6-high, Reasoning: high, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash-high-preview, Reasoning: high, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 34,277 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 23,493 published pairwise votes.
Winner1,522.054
95% confidence interval: ±4.811The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,448.579
95% confidence interval: ±5.563The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlMath
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 4,642 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 2,699 published pairwise votes.
Winner1,505.04
95% confidence interval: ±9.289The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,424.989
95% confidence interval: ±11.731The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlMath
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6-high, Reasoning: high, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash-high-preview, Reasoning: high, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 4,148 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 2,625 published pairwise votes.
Winner1,517.433
95% confidence interval: ±9.697The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,441.363
95% confidence interval: ±11.858The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlMathematical professions
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 4,797 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 2,980 published pairwise votes.
Winner1,512.7
95% confidence interval: ±9.399The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,434.613
95% confidence interval: ±11.415The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlMathematical professions
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6-high, Reasoning: high, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash-high-preview, Reasoning: high, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 4,208 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 2,825 published pairwise votes.
Winner1,525.513
95% confidence interval: ±9.98The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,444.604
95% confidence interval: ±11.746The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlMedicine and healthcare
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 6,382 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 3,692 published pairwise votes.
Winner1,512.795
95% confidence interval: ±8.433The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,459.398
95% confidence interval: ±10.484The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlMedicine and healthcare
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6-high, Reasoning: high, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash-high-preview, Reasoning: high, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 6,092 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 3,760 published pairwise votes.
Winner1,518.12
95% confidence interval: ±8.547The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,461.041
95% confidence interval: ±10.433The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlMulti-turn
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 14,990 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 9,349 published pairwise votes.
Winner1,512.066
95% confidence interval: ±5.994The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,453.168
95% confidence interval: ±7.275The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlMulti-turn
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6-high, Reasoning: high, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash-high-preview, Reasoning: high, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 13,729 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 9,223 published pairwise votes.
Winner1,517.415
95% confidence interval: ±6.216The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,446.213
95% confidence interval: ±7.326The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlNon-English
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 47,281 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 28,879 published pairwise votes.
Winner1,485.029
95% confidence interval: ±3.996The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,422.96
95% confidence interval: ±4.937The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlNon-English
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6-high, Reasoning: high, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash-high-preview, Reasoning: high, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 45,212 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 28,790 published pairwise votes.
Winner1,489.27
95% confidence interval: ±4.076The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,423.369
95% confidence interval: ±5.014The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlPolish
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 1,595 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 1,027 published pairwise votes.
Winner1,494.127
95% confidence interval: ±15.586The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,436.594
95% confidence interval: ±18.719The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlPolish
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6-high, Reasoning: high, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash-high-preview, Reasoning: high, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 1,527 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 1,040 published pairwise votes.
Winner1,501.723
95% confidence interval: ±15.429The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,433.769
95% confidence interval: ±18.399The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlRussian
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 9,657 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 5,766 published pairwise votes.
Winner1,504.697
95% confidence interval: ±6.819The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,432.997
95% confidence interval: ±8.479The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlRussian
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6-high, Reasoning: high, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash-high-preview, Reasoning: high, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 9,274 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 5,793 published pairwise votes.
Winner1,506.054
95% confidence interval: ±6.965The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,436.964
95% confidence interval: ±8.567The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlSoftware and IT services
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 33,350 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 21,384 published pairwise votes.
Winner1,535.811
95% confidence interval: ±4.602The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,471.26
95% confidence interval: ±5.499The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlSoftware and IT services
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6-high, Reasoning: high, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash-high-preview, Reasoning: high, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 30,709 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 21,349 published pairwise votes.
Winner1,538.109
95% confidence interval: ±4.73The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,472.791
95% confidence interval: ±5.525The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlSpanish
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 2,834 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 1,580 published pairwise votes.
Winner1,499.257
95% confidence interval: ±12.441The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,431.488
95% confidence interval: ±16.062The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026
LMArena Text, Style ControlSpanish
Test details
No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6-high, Reasoning: high, Harness: Arena Text, style control
DeepSeek-V4-Flash: Snapshot: deepseek-v4-flash-high-preview, Reasoning: high, Harness: Arena Text, style controlClaude Opus 4.6: Arena rating from 2,745 published pairwise votes.DeepSeek-V4-Flash: Arena rating from 1,476 published pairwise votes.
Winner1,493.263
95% confidence interval: ±12.592The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.
1,429.907
95% confidence interval: ±16.461The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.		LMArena
Source details
Data as of: 10/13/2026

Key to the marks:better valuebetter, but not conclusivetie

Our model tests

Our tests

SVG short test

One SVG prompt and fixed output settings. Captures are shown only with 0 reported reasoning tokens or documented lack of reasoning controls. Identical test conditions do not make the models' internal compute identical.

Shared test round: 4ae5d8eeb9a9

Claude Opus 4.6

Claude Opus 4.6: SVG short test
API reports 0 reasoning tokensTest run: Oct 12, 2026Provider: anthropicAPI model: anthropic/claude-opus-4.6756 output tokens

DeepSeek-V4-Flash

API request failed. The model was not evaluated.

Earlier tests

Pelican on a bicycle

A community classic for free-form SVG drawing.

Claude Opus 4.6

Claude Opus 4.6: Pelican on a bicycle
A community classic for free-form SVG drawing.

Reasoning (requested): high

Current attempt
Run date
Oct 1, 2026
API provider
openrouter
Requested API model
anthropic/claude-opus-4.6
API model reported in response
anthropic/claude-opus-4.6
Reasoning setting
high
Pinned endpoint
Anthropic | anthropic/claude-4.6-opus-20260205
Temperature
Provider default
Token limit including reasoning
8,192 tokens
Test protocol
community-visual-tests-v1
Technical checks
Passed, not a quality rating
Response time
85.1 s

Visually reviewed. This individual test does not establish overall model quality.

DeepSeek-V4-Flash

The response reached the token limit. No complete recording is available.

Reasoning (requested): high

Current attempt
Run date
Oct 5, 2026
API provider
openrouter
Requested API model
deepseek/deepseek-v4-flash-0731
API model reported in response
deepseek/deepseek-v4-flash-0731
Reasoning setting
high
Pinned endpoint
StreamLake | deepseek/deepseek-v4-flash-20260731
Temperature
Provider default
Token limit including reasoning
8,192 tokens
Test protocol
community-visual-tests-v1
Technical checks
Failed
Response time
97.8 s

Not yet visually reviewed. A successful recording does not confirm correct physics or full compliance with the prompt.

Earlier attempts (3)
  • Oct 2, 2026API request failedReasoning setting: high
  • Oct 1, 2026API request failedReasoning setting: high
  • Sep 7, 2026API request failedReasoning setting: high
Prompt and test conditions

Generate an SVG of a pelican riding a bicycle

Token limit including reasoning: 8,192 tokens

Task origin (Simon Willison)

How to read this comparison

Winning one benchmark is not an overall verdict

Your workload, budget, and required context length matter most. A coding benchmark says little about visual understanding or agent performance.

The standard price table uses direct developer rates. The provider comparison labels OpenRouter endpoints, regions, and special tariffs separately.

Every result links to its measurement source and retrieval date. Labels distinguish vendor reports, official benchmark leaderboards, and independent evaluations. Disputed or archived results are excluded.

More matchups

Related LLM comparisons

Compare Claude Opus 4.6 and DeepSeek-V4-Flash with other leading models using the same data and benchmark logic.

View the complete LLM comparison