Head-to-head comparison
Compare published benchmark results from matching versions and cohorts, API pricing, and technical specifications. Any documented test setup differences remain visible.
Specifications
Release date, availability, context window, architecture, and published parameter counts at a glance. Undisclosed details remain marked as unavailable.
| Attribute | Grok 4.6 | DeepSeek-V4-Pro |
|---|---|---|
| Provider | xAI | DeepSeek |
| Released | Aug 12, 2026 | Apr 24, 2026 |
| Availability | Active | Preview |
| Model type | Proprietary | Open Weights |
| Parameters | Not published | 1.6T, 49B active |
| Architecture | Not published | Mixture of Experts |
| Context window | 500,000 tokens | 1,000,000 tokens |
| Knowledge cutoff | February 1, 2026 | Not published |
| Technical sources | Model | Model, Context |
Pricing
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.
| Price type | Grok 4.6 | DeepSeek-V4-Pro |
|---|---|---|
| API input | $2 ≤200K / $4 >200K | $0.66 |
| API output | $6 ≤200K / $12 >200K | $1.98 |
| Cache read | $0.5 ≤200K / $1 >200K | $0.02 |
| Cache write (5 min.) | Not listed | Not listed |
| Cache write (1 hr.) | Not listed | Not listed |
| Price verified | 08/16/2026 Source | 09/08/2026 Source |
Key to the marks:better valuebetter, but not conclusivetie
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.
| Model | Provider | Purchase route and tariff | Input per 1M tokens | Output per 1M tokens | Source |
|---|---|---|---|---|---|
| Grok 4.6 | xAI | Direct from developerUp to 200,000 context tokens | Lowest listed price: $2 | Lowest listed price: $6 | Source08/16/2026 |
| xAI | Direct from developerAbove 200,000 context tokens | $4 | $12 | Source08/16/2026 | |
| Amazon Bedrock | Via OpenRouteramazon-bedrock/us-west-2 | $2.2 | $6.6 | Source09/08/2026 | |
| xAI | Via OpenRouterxai | Lowest listed price: $2 | Lowest listed price: $6 | Source09/08/2026 | |
| xAI | Via OpenRouterxai/priority | $4 | $12 | Source09/08/2026 | |
| xAI | Via OpenRouterxai/zdr | Lowest listed price: $2 | Lowest listed price: $6 | Source09/08/2026 | |
| xAI | Via OpenRouterxai/zdr/priority | $4 | $12 | Source09/08/2026 | |
| DeepSeek-V4-Pro | DeepSeek | Direct from developerStandard API | Lowest listed price: $0.66 | Lowest listed price: $1.98 | Source09/08/2026 |
| Alibaba | Via OpenRouteralibaba | $1.122 | $3.366 | Source09/08/2026 | |
| Baidu | Via OpenRouterbaidu/fp8 | $1.1194 | $3.3581 | Source09/08/2026 | |
| BaseTen | Via OpenRouterbaseten/fp4#1 | $1.32 | $3.96 | Source09/08/2026 | |
| BaseTen | Via OpenRouterbaseten/fp4#2 | $1.32 | $3.96 | Source09/08/2026 | |
| Cloudflare | Via OpenRoutercloudflare | $1.32 | $3.96 | Source09/08/2026 | |
| CoreWeave | Via OpenRoutercoreweave/fp8 | $1.31 | $3.96 | Source09/08/2026 | |
| DeepInfra | Via OpenRouterdeepinfra/fp8 | $1.3 | $2.6 | Source09/08/2026 | |
| DeepSeek | Via OpenRouterdeepseek | Lowest listed price: $0.66 | $1.98 | Source09/08/2026 | |
| DigitalOcean | Via OpenRouterdigitalocean | $1.32 | $3.96 | Source09/08/2026 | |
| Fireworks | Via OpenRouterfireworks | $1.32 | $3.96 | Source09/08/2026 | |
| GMICloud | Via OpenRoutergmicloud/fp8 | $1.056 | $3.168 | Source09/08/2026 | |
| Ionstream | Via OpenRouterionstream/fp4 | $1.1 | $3.3 | Source09/08/2026 | |
| NextBit | Via OpenRouternextbit/fp8 | $1.122 | $3.366 | Source09/08/2026 | |
| Novita | Via OpenRouternovita/fp8 | $0.99 | $2.97 | Source09/08/2026 | |
| Parasail | Via OpenRouterparasail/fp8 | $1.32 | $3.96 | Source09/08/2026 | |
| Phala | Via OpenRouterphala | $1.45 | $4.36 | Source09/08/2026 | |
| Sail Research | Via OpenRoutersail-research/fp4 | $1.32 | $3.96 | Source09/08/2026 | |
| SiliconFlow | Via OpenRoutersiliconflow/fp8 | $1.32 | $3.96 | Source09/08/2026 | |
| StreamLake | Via OpenRouterstreamlake | $1.0494 | $3.1482 | Source09/08/2026 | |
| Together | Via OpenRoutertogether | $1.32 | $3.96 | Source09/08/2026 |
Performance
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.
Each axis averages directly matched benchmark families. A value of 100 means the stronger value within this pair, not a universal quality score.
43 matched rows from 21 benchmark families were considered. The radar shows 8 of 11 comparable categories.
Key to the marks:better valuebetter, but not conclusivetie
| Benchmark and task | Grok 4.6 | DeepSeek-V4-Pro | Source |
|---|---|---|---|
Code MigrationOverallTest detailsNo documented difference in the test setupVersion: 1Metric: accuracyScoring: Higher is better | Winner44.572 % | 26.201 % | Vals AISource detailsData as of: 09/03/2026 |
CorpFin v2OverallTest detailsNo documented difference in the test setupVersion: 2Metric: accuracyScoring: Higher is better | Winner66.161 % | 61.383 % | Vals AISource detailsData as of: 08/12/2026 |
EMBOverallTest detailsNo documented difference in the test setupVersion: 1Metric: accuracyScoring: Higher is better | Winner62.727 % | 51.623 % | Vals AISource detailsData as of: 09/03/2026 |
Finance Agent (v2)OverallTest detailsNo documented difference in the test setupVersion: 2Metric: accuracyScoring: Higher is better | Winner53.682 % | 44.083 % | Vals AISource detailsData as of: 09/03/2026 |
GPQA DiamondOverallTest detailsNo documented difference in the test setupVersion: 1Metric: accuracyScoring: Higher is better | Winner94.697 % | 89.394 % | Vals AISource detailsData as of: 09/01/2026 |
Harvey's Legal Agent BenchmarkOverall · Task fully resolvedTest detailsNo documented difference in the test setupVersion: 1Metric: task resolution rateScoring: Higher is better | Winner15.833 % | 3.75 % | Vals AISource detailsData as of: 09/03/2026 |
Legal Research BenchOverall · All-passTest detailsNo documented difference in the test setupVersion: 1Metric: all-pass rateScoring: Higher is better | Winner48.077 % | 23.077 % | Vals AISource detailsData as of: 09/03/2026 |
LegalBenchOverallTest detailsNo documented difference in the test setupVersion: 1Metric: accuracyScoring: Higher is better | Winner86.307 % | 80.323 % | Vals AISource detailsData as of: 09/01/2026 |
LiveCodeBench v6OverallTest detailsNo documented difference in the test setupVersion: 1Metric: accuracyScoring: Higher is better | Winner88.224 % | 87.484 % | Vals AISource detailsData as of: 09/01/2026 |
LMArena Text, Style ControlOverallTest detailsTest setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGrok 4.6: Snapshot: grok-4.6-high, Reasoning: high, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro, Harness: Arena text, style controlGrok 4.6: Style-controlled Arena rating from 3,453 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 54,243 anonymous pairwise votes. | Higher measured value1,461.115 95% confidence interval: ±10.144The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | 1,457.748 95% confidence interval: ±4.009The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | LMArenaSource detailsData as of: 09/01/2026 |
MedCodeOverallTest detailsNo documented difference in the test setupVersion: 1Metric: accuracyScoring: Higher is better | Winner44.712 % | 40.455 % | Vals AISource detailsData as of: 09/03/2026 |
MedScribeOverallTest detailsNo documented difference in the test setupVersion: 1Metric: accuracyScoring: Higher is better | Winner86.534 % | 75.144 % | Vals AISource detailsData as of: 09/03/2026 |
MMLU-ProOverallTest detailsNo documented difference in the test setupVersion: 1Metric: accuracyScoring: Higher is better | Winner89.4 % | 87.249 % | Vals AISource detailsData as of: 09/01/2026 |
ProofBenchOverallTest detailsNo documented difference in the test setupVersion: 1.1Metric: accuracyScoring: Higher is better | Winner51 % | 16 % | Vals AISource detailsData as of: 09/01/2026 |
Public Benefits Bench v1.1OverallTest detailsNo documented difference in the test setupVersion: 1.1Metric: accuracyScoring: Higher is better | Winner66.847 % | 62.923 % | Vals AISource detailsData as of: 09/01/2026 |
SkillsBenchOverallTest detailsNo documented difference in the test setupVersion: 1Metric: accuracyScoring: Higher is betterSettings: Harness: OpenHands | Winner55.766 % | 51.274 % | Vals AISource detailsData as of: 09/01/2026 |
SWE-bench VerifiedOverallTest detailsNo documented difference in the test setupVersion: 1Metric: accuracyScoring: Higher is better | Winner95.6 % | 77.4 % | Vals AISource detailsData as of: 09/01/2026 |
TaxEval v2OverallTest detailsNo documented difference in the test setupVersion: 2Metric: accuracyScoring: Higher is better | 71.096 % | Winner72.077 % | Vals AISource detailsData as of: 09/01/2026 |
Terminal-Bench 2.1OverallTest detailsNo documented difference in the test setupVersion: 2.1Metric: accuracyScoring: Higher is better | Winner78.277 % | 50.187 % | Vals AISource detailsData as of: 09/03/2026 |
Vals IndexOverallTest detailsNo documented difference in the test setupVersion: 2Metric: weighted index scoreScoring: Higher is better | Winner59.167 % | 42.888 % | Vals AISource detailsData as of: 09/03/2026 |
Vibe Code Bench v1.1OverallTest detailsNo documented difference in the test setupVersion: 1.1Metric: accuracyScoring: Higher is betterSettings: Harness: OpenHands | Winner76.239 % | 49.931 % | Vals AISource detailsData as of: 09/03/2026 |
LMArena Text, Style ControlOverallTest detailsNo documented difference in the test setupNo clear leadVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGrok 4.6: Snapshot: grok-4.6-high, Reasoning: high, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro-high-preview, Reasoning: high, Harness: Arena text, style controlGrok 4.6: Style-controlled Arena rating from 3,453 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 51,599 anonymous pairwise votes. | No clear lead1,461.115 95% confidence interval: ±10.144The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | 1,455.047 95% confidence interval: ±4.08The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | LMArenaSource detailsData as of: 09/01/2026 |
LMArena Text, Style ControlBusiness, management and financeTest detailsTest setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGrok 4.6: Snapshot: grok-4.6-high, Reasoning: high, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro, Harness: Arena text, style controlGrok 4.6: Style-controlled Arena rating from 705 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 10,742 anonymous pairwise votes. | Higher measured value1,467.318 95% confidence interval: ±22.146The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | 1,461.034 95% confidence interval: ±6.908The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | LMArenaSource detailsData as of: 09/01/2026 |
LMArena Text, Style ControlBusiness, management and financeTest detailsNo documented difference in the test setupNo clear leadVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGrok 4.6: Snapshot: grok-4.6-high, Reasoning: high, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro-high-preview, Reasoning: high, Harness: Arena text, style controlGrok 4.6: Style-controlled Arena rating from 705 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 10,411 anonymous pairwise votes. | No clear lead1,467.318 95% confidence interval: ±22.146The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | 1,448.251 95% confidence interval: ±6.986The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | LMArenaSource detailsData as of: 09/01/2026 |
LMArena Text, Style ControlChineseTest detailsTest setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGrok 4.6: Snapshot: grok-4.6-high, Reasoning: high, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro, Harness: Arena text, style controlGrok 4.6: Style-controlled Arena rating from 291 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 2,871 anonymous pairwise votes. | Higher measured value1,538.386 95% confidence interval: ±33.69The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | 1,493.385 95% confidence interval: ±11.839The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | LMArenaSource detailsData as of: 09/01/2026 |
LMArena Text, Style ControlChineseTest detailsNo documented difference in the test setupNo clear leadVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGrok 4.6: Snapshot: grok-4.6-high, Reasoning: high, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro-high-preview, Reasoning: high, Harness: Arena text, style controlGrok 4.6: Style-controlled Arena rating from 291 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 2,757 anonymous pairwise votes. | No clear lead1,538.386 95% confidence interval: ±33.69The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | 1,498.025 95% confidence interval: ±12.143The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | LMArenaSource detailsData as of: 09/01/2026 |
LMArena Text, Style ControlCodingTest detailsTest setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGrok 4.6: Snapshot: grok-4.6-high, Reasoning: high, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro, Harness: Arena text, style controlGrok 4.6: Style-controlled Arena rating from 838 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 16,165 anonymous pairwise votes. | Higher measured value1,510.751 95% confidence interval: ±20.826The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | 1,501.769 95% confidence interval: ±6.076The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | LMArenaSource detailsData as of: 09/01/2026 |
LMArena Text, Style ControlCodingTest detailsNo documented difference in the test setupNo clear leadVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGrok 4.6: Snapshot: grok-4.6-high, Reasoning: high, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro-high-preview, Reasoning: high, Harness: Arena text, style controlGrok 4.6: Style-controlled Arena rating from 838 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 15,113 anonymous pairwise votes. | No clear lead1,510.751 95% confidence interval: ±20.826The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | 1,488.503 95% confidence interval: ±6.226The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | LMArenaSource detailsData as of: 09/01/2026 |
LMArena Text, Style ControlCreative writingTest detailsTest setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGrok 4.6: Snapshot: grok-4.6-high, Reasoning: high, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro, Harness: Arena text, style controlGrok 4.6: Style-controlled Arena rating from 651 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 9,051 anonymous pairwise votes. | Higher measured value1,458.096 95% confidence interval: ±23.926The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | 1,444.463 95% confidence interval: ±7.495The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | LMArenaSource detailsData as of: 09/01/2026 |
LMArena Text, Style ControlCreative writingTest detailsNo documented difference in the test setupNo clear leadVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGrok 4.6: Snapshot: grok-4.6-high, Reasoning: high, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro-high-preview, Reasoning: high, Harness: Arena text, style controlGrok 4.6: Style-controlled Arena rating from 651 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 8,660 anonymous pairwise votes. | No clear lead1,458.096 95% confidence interval: ±23.926The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | 1,440.457 95% confidence interval: ±7.584The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | LMArenaSource detailsData as of: 09/01/2026 |
LMArena Text, Style ControlEnglishTest detailsTest setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGrok 4.6: Snapshot: grok-4.6-high, Reasoning: high, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro, Harness: Arena text, style controlGrok 4.6: Style-controlled Arena rating from 1,434 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 23,924 anonymous pairwise votes. | 1,459.46 95% confidence interval: ±15.407The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | Higher measured value1,467.254 95% confidence interval: ±5.227The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | LMArenaSource detailsData as of: 09/01/2026 |
LMArena Text, Style ControlEnglishTest detailsNo documented difference in the test setupNo clear leadVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGrok 4.6: Snapshot: grok-4.6-high, Reasoning: high, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro-high-preview, Reasoning: high, Harness: Arena text, style controlGrok 4.6: Style-controlled Arena rating from 1,434 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 22,700 anonymous pairwise votes. | 1,459.46 95% confidence interval: ±15.407The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | No clear lead1,465.768 95% confidence interval: ±5.28The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | LMArenaSource detailsData as of: 09/01/2026 |
LMArena Text, Style ControlEntertainment, sports and mediaTest detailsTest setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGrok 4.6: Snapshot: grok-4.6-high, Reasoning: high, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro, Harness: Arena text, style controlGrok 4.6: Style-controlled Arena rating from 777 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 11,880 anonymous pairwise votes. | Higher measured value1,443.149 95% confidence interval: ±21.821The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | 1,433.955 95% confidence interval: ±6.804The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | LMArenaSource detailsData as of: 09/01/2026 |
LMArena Text, Style ControlEntertainment, sports and mediaTest detailsNo documented difference in the test setupNo clear leadVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGrok 4.6: Snapshot: grok-4.6-high, Reasoning: high, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro-high-preview, Reasoning: high, Harness: Arena text, style controlGrok 4.6: Style-controlled Arena rating from 777 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 11,440 anonymous pairwise votes. | No clear lead1,443.149 95% confidence interval: ±21.821The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | 1,430.064 95% confidence interval: ±6.877The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | LMArenaSource detailsData as of: 09/01/2026 |
LMArena Text, Style ControlExcluding tiesTest detailsTest setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGrok 4.6: Snapshot: grok-4.6-high, Reasoning: high, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro, Harness: Arena text, style controlGrok 4.6: Style-controlled Arena rating from 2,522 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 41,089 anonymous pairwise votes. | Higher measured value1,465.128 95% confidence interval: ±14.002The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | 1,461.499 95% confidence interval: ±5.248The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | LMArenaSource detailsData as of: 09/01/2026 |
LMArena Text, Style ControlExcluding tiesTest detailsNo documented difference in the test setupNo clear leadVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGrok 4.6: Snapshot: grok-4.6-high, Reasoning: high, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro-high-preview, Reasoning: high, Harness: Arena text, style controlGrok 4.6: Style-controlled Arena rating from 2,522 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 39,187 anonymous pairwise votes. | No clear lead1,465.128 95% confidence interval: ±14.002The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | 1,457.114 95% confidence interval: ±5.263The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | LMArenaSource detailsData as of: 09/01/2026 |
LMArena Text, Style ControlExpert promptsTest detailsTest setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGrok 4.6: Snapshot: grok-4.6-high, Reasoning: high, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro, Harness: Arena text, style controlGrok 4.6: Style-controlled Arena rating from 401 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 5,439 anonymous pairwise votes. | 1,479.844 95% confidence interval: ±28.841The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | Higher measured value1,481.602 95% confidence interval: ±8.815The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | LMArenaSource detailsData as of: 09/01/2026 |
LMArena Text, Style ControlExpert promptsTest detailsNo documented difference in the test setupNo clear leadVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGrok 4.6: Snapshot: grok-4.6-high, Reasoning: high, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro-high-preview, Reasoning: high, Harness: Arena text, style controlGrok 4.6: Style-controlled Arena rating from 401 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 5,063 anonymous pairwise votes. | No clear lead1,479.844 95% confidence interval: ±28.841The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | 1,475.253 95% confidence interval: ±8.974The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | LMArenaSource detailsData as of: 09/01/2026 |
LMArena Text, Style ControlHard promptsTest detailsTest setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGrok 4.6: Snapshot: grok-4.6-high, Reasoning: high, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro, Harness: Arena text, style controlGrok 4.6: Style-controlled Arena rating from 2,312 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 36,098 anonymous pairwise votes. | Higher measured value1,484.219 95% confidence interval: ±12.5The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | 1,480.273 95% confidence interval: ±4.693The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | LMArenaSource detailsData as of: 09/01/2026 |
LMArena Text, Style ControlHard promptsTest detailsNo documented difference in the test setupNo clear leadVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGrok 4.6: Snapshot: grok-4.6-high, Reasoning: high, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro-high-preview, Reasoning: high, Harness: Arena text, style controlGrok 4.6: Style-controlled Arena rating from 2,312 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 34,551 anonymous pairwise votes. | No clear lead1,484.219 95% confidence interval: ±12.5The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | 1,474.049 95% confidence interval: ±4.754The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | LMArenaSource detailsData as of: 09/01/2026 |
LMArena Text, Style ControlHard prompts, EnglishTest detailsTest setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGrok 4.6: Snapshot: grok-4.6-high, Reasoning: high, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro, Harness: Arena text, style controlGrok 4.6: Style-controlled Arena rating from 927 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 16,819 anonymous pairwise votes. | 1,475.235 95% confidence interval: ±19.394The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | Higher measured value1,484.473 95% confidence interval: ±6.001The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | LMArenaSource detailsData as of: 09/01/2026 |
LMArena Text, Style ControlHard prompts, EnglishTest detailsNo documented difference in the test setupNo clear leadVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGrok 4.6: Snapshot: grok-4.6-high, Reasoning: high, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro-high-preview, Reasoning: high, Harness: Arena text, style controlGrok 4.6: Style-controlled Arena rating from 927 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 15,908 anonymous pairwise votes. | 1,475.235 95% confidence interval: ±19.394The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | No clear lead1,482.359 95% confidence interval: ±6.095The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | LMArenaSource detailsData as of: 09/01/2026 |
LMArena Text, Style ControlInstruction followingTest detailsTest setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGrok 4.6: Snapshot: grok-4.6-high, Reasoning: high, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro, Harness: Arena text, style controlGrok 4.6: Style-controlled Arena rating from 1,223 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 18,672 anonymous pairwise votes. | Higher measured value1,458.276 95% confidence interval: ±17.043The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | 1,453.345 95% confidence interval: ±5.73The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | LMArenaSource detailsData as of: 09/01/2026 |
LMArena Text, Style ControlInstruction followingTest detailsNo documented difference in the test setupNo clear leadVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGrok 4.6: Snapshot: grok-4.6-high, Reasoning: high, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro-high-preview, Reasoning: high, Harness: Arena text, style controlGrok 4.6: Style-controlled Arena rating from 1,223 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 17,742 anonymous pairwise votes. | No clear lead1,458.276 95% confidence interval: ±17.043The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | 1,449.073 95% confidence interval: ±5.821The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | LMArenaSource detailsData as of: 09/01/2026 |
LMArena Text, Style ControlLegal and governmentTest detailsTest setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGrok 4.6: Snapshot: grok-4.6-high, Reasoning: high, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro, Harness: Arena text, style controlGrok 4.6: Style-controlled Arena rating from 284 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 4,235 anonymous pairwise votes. | Higher measured value1,475.664 95% confidence interval: ±33.128The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | 1,473.424 95% confidence interval: ±9.866The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | LMArenaSource detailsData as of: 09/01/2026 |
LMArena Text, Style ControlLegal and governmentTest detailsNo documented difference in the test setupNo clear leadVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGrok 4.6: Snapshot: grok-4.6-high, Reasoning: high, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro-high-preview, Reasoning: high, Harness: Arena text, style controlGrok 4.6: Style-controlled Arena rating from 284 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 4,061 anonymous pairwise votes. | No clear lead1,475.664 95% confidence interval: ±33.128The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | 1,471.913 95% confidence interval: ±10.024The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | LMArenaSource detailsData as of: 09/01/2026 |
LMArena Text, Style ControlLife, physical and social scienceTest detailsTest setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGrok 4.6: Snapshot: grok-4.6-high, Reasoning: high, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro, Harness: Arena text, style controlGrok 4.6: Style-controlled Arena rating from 512 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 8,607 anonymous pairwise votes. | Higher measured value1,492.359 95% confidence interval: ±25.684The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | 1,480.819 95% confidence interval: ±7.337The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | LMArenaSource detailsData as of: 09/01/2026 |
LMArena Text, Style ControlLife, physical and social scienceTest detailsNo documented difference in the test setupNo clear leadVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGrok 4.6: Snapshot: grok-4.6-high, Reasoning: high, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro-high-preview, Reasoning: high, Harness: Arena text, style controlGrok 4.6: Style-controlled Arena rating from 512 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 8,161 anonymous pairwise votes. | No clear lead1,492.359 95% confidence interval: ±25.684The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | 1,478.979 95% confidence interval: ±7.429The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | LMArenaSource detailsData as of: 09/01/2026 |
LMArena Text, Style ControlLonger queriesTest detailsTest setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGrok 4.6: Snapshot: grok-4.6-high, Reasoning: high, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro, Harness: Arena text, style controlGrok 4.6: Style-controlled Arena rating from 1,609 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 23,973 anonymous pairwise votes. | Higher measured value1,478.218 95% confidence interval: ±15.132The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | 1,472.211 95% confidence interval: ±5.5The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | LMArenaSource detailsData as of: 09/01/2026 |
LMArena Text, Style ControlLonger queriesTest detailsNo documented difference in the test setupNo clear leadVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGrok 4.6: Snapshot: grok-4.6-high, Reasoning: high, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro-high-preview, Reasoning: high, Harness: Arena text, style controlGrok 4.6: Style-controlled Arena rating from 1,609 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 22,832 anonymous pairwise votes. | No clear lead1,478.218 95% confidence interval: ±15.132The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | 1,466.143 95% confidence interval: ±5.577The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | LMArenaSource detailsData as of: 09/01/2026 |
LMArena Text, Style ControlMathTest detailsTest setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGrok 4.6: Snapshot: grok-4.6-high, Reasoning: high, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro, Harness: Arena text, style controlGrok 4.6: Style-controlled Arena rating from 216 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 2,868 anonymous pairwise votes. | 1,426.012 95% confidence interval: ±40.021The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | Higher measured value1,444.594 95% confidence interval: ±11.525The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | LMArenaSource detailsData as of: 09/01/2026 |
LMArena Text, Style ControlMathTest detailsNo documented difference in the test setupNo clear leadVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGrok 4.6: Snapshot: grok-4.6-high, Reasoning: high, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro-high-preview, Reasoning: high, Harness: Arena text, style controlGrok 4.6: Style-controlled Arena rating from 216 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 2,505 anonymous pairwise votes. | 1,426.012 95% confidence interval: ±40.021The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | No clear lead1,469.012 95% confidence interval: ±12.249The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | LMArenaSource detailsData as of: 09/01/2026 |
LMArena Text, Style ControlMathematical professionsTest detailsTest setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGrok 4.6: Snapshot: grok-4.6-high, Reasoning: high, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro, Harness: Arena text, style controlGrok 4.6: Style-controlled Arena rating from 212 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 3,010 anonymous pairwise votes. | 1,456.037 95% confidence interval: ±39.211The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | Higher measured value1,460.773 95% confidence interval: ±11.504The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | LMArenaSource detailsData as of: 09/01/2026 |
LMArena Text, Style ControlMathematical professionsTest detailsNo documented difference in the test setupNo clear leadVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGrok 4.6: Snapshot: grok-4.6-high, Reasoning: high, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro-high-preview, Reasoning: high, Harness: Arena text, style controlGrok 4.6: Style-controlled Arena rating from 212 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 2,693 anonymous pairwise votes. | 1,456.037 95% confidence interval: ±39.211The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | No clear lead1,469.984 95% confidence interval: ±12.095The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | LMArenaSource detailsData as of: 09/01/2026 |
LMArena Text, Style ControlMedicine and healthcareTest detailsTest setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGrok 4.6: Snapshot: grok-4.6-high, Reasoning: high, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro, Harness: Arena text, style controlGrok 4.6: Style-controlled Arena rating from 225 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 3,762 anonymous pairwise votes. | 1,460.477 95% confidence interval: ±37.871The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | Higher measured value1,477.327 95% confidence interval: ±10.549The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | LMArenaSource detailsData as of: 09/01/2026 |
LMArena Text, Style ControlMedicine and healthcareTest detailsNo documented difference in the test setupNo clear leadVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGrok 4.6: Snapshot: grok-4.6-high, Reasoning: high, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro-high-preview, Reasoning: high, Harness: Arena text, style controlGrok 4.6: Style-controlled Arena rating from 225 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 3,584 anonymous pairwise votes. | 1,460.477 95% confidence interval: ±37.871The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | No clear lead1,485.1 95% confidence interval: ±10.705The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | LMArenaSource detailsData as of: 09/01/2026 |
LMArena Text, Style ControlMulti-turnTest detailsTest setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGrok 4.6: Snapshot: grok-4.6-high, Reasoning: high, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro, Harness: Arena text, style controlGrok 4.6: Style-controlled Arena rating from 569 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 9,810 anonymous pairwise votes. | 1,446.12 95% confidence interval: ±25.079The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | Higher measured value1,475.824 95% confidence interval: ±7.2The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | LMArenaSource detailsData as of: 09/01/2026 |
LMArena Text, Style ControlMulti-turnTest detailsNo documented difference in the test setupNo clear leadVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGrok 4.6: Snapshot: grok-4.6-high, Reasoning: high, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro-high-preview, Reasoning: high, Harness: Arena text, style controlGrok 4.6: Style-controlled Arena rating from 569 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 9,139 anonymous pairwise votes. | 1,446.12 95% confidence interval: ±25.079The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | No clear lead1,455.379 95% confidence interval: ±7.327The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | LMArenaSource detailsData as of: 09/01/2026 |
LMArena Text, Style ControlNon-EnglishTest detailsTest setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGrok 4.6: Snapshot: grok-4.6-high, Reasoning: high, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro, Harness: Arena text, style controlGrok 4.6: Style-controlled Arena rating from 2,019 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 30,319 anonymous pairwise votes. | Higher measured value1,456.123 95% confidence interval: ±13.128The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | 1,445.037 95% confidence interval: ±4.83The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | LMArenaSource detailsData as of: 09/01/2026 |
LMArena Text, Style ControlNon-EnglishTest detailsNo documented difference in the test setupNo clear leadVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGrok 4.6: Snapshot: grok-4.6-high, Reasoning: high, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro-high-preview, Reasoning: high, Harness: Arena text, style controlGrok 4.6: Style-controlled Arena rating from 2,019 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 28,895 anonymous pairwise votes. | No clear lead1,456.123 95% confidence interval: ±13.128The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | 1,440.825 95% confidence interval: ±4.904The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | LMArenaSource detailsData as of: 09/01/2026 |
LMArena Text, Style ControlRussianTest detailsTest setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGrok 4.6: Snapshot: grok-4.6-high, Reasoning: high, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro, Harness: Arena text, style controlGrok 4.6: Style-controlled Arena rating from 413 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 5,710 anonymous pairwise votes. | 1,446.297 95% confidence interval: ±28.9The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | Higher measured value1,458.01 95% confidence interval: ±8.611The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | LMArenaSource detailsData as of: 09/01/2026 |
LMArena Text, Style ControlRussianTest detailsNo documented difference in the test setupNo clear leadVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGrok 4.6: Snapshot: grok-4.6-high, Reasoning: high, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro-high-preview, Reasoning: high, Harness: Arena text, style controlGrok 4.6: Style-controlled Arena rating from 413 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 5,539 anonymous pairwise votes. | 1,446.297 95% confidence interval: ±28.9The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | No clear lead1,458.648 95% confidence interval: ±8.697The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | LMArenaSource detailsData as of: 09/01/2026 |
LMArena Text, Style ControlSoftware and IT servicesTest detailsTest setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGrok 4.6: Snapshot: grok-4.6-high, Reasoning: high, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro, Harness: Arena text, style controlGrok 4.6: Style-controlled Arena rating from 1,379 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 22,292 anonymous pairwise votes. | Higher measured value1,503.685 95% confidence interval: ±16.15The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | 1,491.975 95% confidence interval: ±5.393The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | LMArenaSource detailsData as of: 09/01/2026 |
LMArena Text, Style ControlSoftware and IT servicesTest detailsNo documented difference in the test setupNo clear leadVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGrok 4.6: Snapshot: grok-4.6-high, Reasoning: high, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro-high-preview, Reasoning: high, Harness: Arena text, style controlGrok 4.6: Style-controlled Arena rating from 1,379 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 21,173 anonymous pairwise votes. | No clear lead1,503.685 95% confidence interval: ±16.15The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | 1,483.273 95% confidence interval: ±5.481The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | LMArenaSource detailsData as of: 09/01/2026 |
Key to the marks:better valuebetter, but not conclusivetie | |||
Our model tests
Compare how the models respond to the same prompt. Each test shows the first attempt, with no subsequent fixes to the generated code. These results do not contribute to an overall score.
A community classic for free-form SVG drawing.
Grok 4.6
API access was unavailable for this run. The model was not evaluated.
Not yet visually reviewed. A successful recording does not confirm correct physics or full compliance with the prompt.
DeepSeek-V4-Pro
API access was unavailable for this run. The model was not evaluated.
Not yet visually reviewed. A successful recording does not confirm correct physics or full compliance with the prompt.
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
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
Compare Grok 4.6 and DeepSeek-V4-Pro with other leading models using the same data and benchmark logic.