Compare published benchmark results from matching versions and cohorts, API pricing, and technical specifications. Any documented test setup differences remain visible.
shared result series
79
distinct benchmarks
22
latest retrieval
2026-09-29
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.
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.
DeepSeek-V4-Pro: peak pricing is $1.32 input / $3.96 output per 1M tokens during Mon, Tue, Wed, Thu, Fri, 01:00-04:00 UTC; Mon, Tue, Wed, Thu, Fri, 06:00-10:00 UTC; off-peak rates apply at all other times.
Grok 4.6DeepSeek-V4-Pro
API inputUSD per 1M tokens, base tier
Grok 4.6$2
DeepSeek-V4-ProLower price$0.66
API outputUSD per 1M tokens, base tier
Grok 4.6$6
DeepSeek-V4-ProLower price$1.98
Cache readUSD per 1M tokens, base tier
Grok 4.6$0.5
DeepSeek-V4-ProLower price$0.022
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.
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.
Grok 4.6
DeepSeek-V4-Pro
Coding: 100 / 80.6 (5 benchmark families)
Reasoning: 99.9 / 97 (2 benchmark families)
Mathematics: 99.6 / 65.4 (2 benchmark families)
Knowledge: 100 / 98.6 (2 benchmark families)
Finance: 99.8 / 90 (6 benchmark families)
Legal: 99.6 / 66.2 (4 benchmark families)
Agentic tasks: 100 / 78 (2 benchmark families)
Productivity: 100 / 94.1 (1 benchmark family)
50 matched rows from 22 benchmark families were considered. The radar shows 8 of 11 comparable categories.
Grok 4.6DeepSeek-V4-Pro
Code MigrationOverall. Higher is better. No documented setup difference
Grok 4.6Winner44.572 %
DeepSeek-V4-Pro26.201 %
CorpFin v2Overall. Higher is better. No documented setup difference
Grok 4.6Winner66.161 %
DeepSeek-V4-Pro61.383 %
EMBOverall. Higher is better. No documented setup difference
Grok 4.6Winner62.727 %
DeepSeek-V4-Pro51.623 %
Finance Agent (v2)Overall. Higher is better. No documented setup difference
Grok 4.6Winner53.682 %
DeepSeek-V4-Pro44.083 %
GPQA DiamondOverall. Higher is better. No documented setup difference
Grok 4.6Winner94.697 %
DeepSeek-V4-Pro89.394 %
Harvey's Legal Agent BenchmarkOverall · Task fully resolved. Higher is better. No documented setup difference
Grok 4.6Winner15.833 %
DeepSeek-V4-Pro3.75 %
Legal Research BenchOverall · All-pass. Higher is better. No documented setup difference
Grok 4.6Winner48.077 %
DeepSeek-V4-Pro23.077 %
LegalBenchOverall. Higher is better. No documented setup difference
Grok 4.6Winner86.307 %
DeepSeek-V4-Pro80.323 %
LiveCodeBench v6Overall. Higher is better. No documented setup difference
Grok 4.6Winner88.224 %
DeepSeek-V4-Pro87.484 %
LMArena Text, Style ControlOverall. Higher is better. Test setup not documented as identical
Grok 4.61,453.397
DeepSeek-V4-ProHigher measured value1,457.519
MedCodeOverall. Higher is better. No documented setup difference
Grok 4.6Winner44.712 %
DeepSeek-V4-Pro40.455 %
MedScribeOverall. Higher is better. No documented setup difference
Grok 4.6Winner86.534 %
DeepSeek-V4-Pro75.144 %
Key to the marks:better valuebetter, but not conclusivetie
Benchmark scores for Grok 4.6 and DeepSeek-V4-Pro
Benchmark and task
Grok 4.6
DeepSeek-V4-Pro
Source
Code MigrationOverallTest details
No documented difference in the test setupVersion: 1Metric: accuracyScoring: Higher is better
Test 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 21,380 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 57,564 anonymous pairwise votes.
1,453.397
95% confidence interval: ±5.242The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
Higher measured value1,457.519
95% confidence interval: ±3.905The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
No 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 21,380 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 54,884 anonymous pairwise votes.
1,453.397
95% confidence interval: ±5.242The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
No clear lead1,454.675
95% confidence interval: ±3.973The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
LMArena Text, Style ControlBusiness, management and financeTest details
Test 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 4,164 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 11,411 anonymous pairwise votes.
1,447.416
95% confidence interval: ±9.612The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
Higher measured value1,459.704
95% confidence interval: ±6.73The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
LMArena Text, Style ControlBusiness, management and financeTest details
No 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 4,164 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 11,076 anonymous pairwise votes.
1,447.416
95% confidence interval: ±9.612The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
No clear lead1,448.077
95% confidence interval: ±6.825The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
Test 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 1,707 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 3,408 anonymous pairwise votes.
Higher measured value1,502.981
95% confidence interval: ±14.739The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,488.151
95% confidence interval: ±10.761The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
No 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 1,707 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 3,360 anonymous pairwise votes.
No clear lead1,502.981
95% confidence interval: ±14.739The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,499.718
95% confidence interval: ±10.912The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
Test 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 5,539 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 17,165 anonymous pairwise votes.
Higher measured value1,507.402
95% confidence interval: ±8.493The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,500.462
95% confidence interval: ±5.93The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
No documented difference in the test setupVersion: 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 5,539 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 16,030 anonymous pairwise votes.
Winner1,507.402
95% confidence interval: ±8.493The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,487.781
95% confidence interval: ±6.099The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
Test 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 4,559 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 9,697 anonymous pairwise votes.
1,444.507
95% confidence interval: ±9.71The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
Higher measured value1,445.978
95% confidence interval: ±7.296The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
No 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 4,559 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 9,329 anonymous pairwise votes.
No clear lead1,444.507
95% confidence interval: ±9.71The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,440.486
95% confidence interval: ±7.37The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
Test 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 8,594 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 25,176 anonymous pairwise votes.
1,456.489
95% confidence interval: ±7.091The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
Higher measured value1,467.134
95% confidence interval: ±5.13The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
No 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 8,594 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 23,891 anonymous pairwise votes.
1,456.489
95% confidence interval: ±7.091The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
No clear lead1,466.543
95% confidence interval: ±5.188The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
LMArena Text, Style ControlEntertainment, sports and mediaTest details
Test 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 5,466 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 12,738 anonymous pairwise votes.
Higher measured value1,440.971
95% confidence interval: ±8.913The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,434.725
95% confidence interval: ±6.63The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
LMArena Text, Style ControlEntertainment, sports and mediaTest details
No 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 5,466 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 12,346 anonymous pairwise votes.
No clear lead1,440.971
95% confidence interval: ±8.913The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,431.889
95% confidence interval: ±6.685The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
Test 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 15,795 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 43,577 anonymous pairwise votes.
1,455.675
95% confidence interval: ±6.93The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
Higher measured value1,461.31
95% confidence interval: ±5.137The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
No 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 15,795 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 41,744 anonymous pairwise votes.
1,455.675
95% confidence interval: ±6.93The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
No clear lead1,456.705
95% confidence interval: ±5.153The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
Test 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 2,489 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 5,925 anonymous pairwise votes.
Higher measured value1,489.565
95% confidence interval: ±12.215The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,479.666
95% confidence interval: ±8.43The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
No 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 2,489 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 5,532 anonymous pairwise votes.
No clear lead1,489.565
95% confidence interval: ±12.215The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,475.361
95% confidence interval: ±8.612The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
Test 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 574 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 2,215 anonymous pairwise votes.
Higher measured value1,488.187
95% confidence interval: ±25.71The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,485.521
95% confidence interval: ±14.661The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
No 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 574 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 2,075 anonymous pairwise votes.
No clear lead1,488.187
95% confidence interval: ±25.71The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,469.327
95% confidence interval: ±14.782The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
Test 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 387 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 1,001 anonymous pairwise votes.
1,455.753
95% confidence interval: ±30.673The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
Higher measured value1,468.915
95% confidence interval: ±19.454The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
No 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 387 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 928 anonymous pairwise votes.
1,455.753
95% confidence interval: ±30.673The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
No clear lead1,460.563
95% confidence interval: ±20.112The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
Test 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 14,049 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 38,586 anonymous pairwise votes.
1,478.04
95% confidence interval: ±6.111The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
Higher measured value1,479.844
95% confidence interval: ±4.586The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
No 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 14,049 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 36,977 anonymous pairwise votes.
No clear lead1,478.04
95% confidence interval: ±6.111The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,473.564
95% confidence interval: ±4.651The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
Test 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 5,649 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 17,828 anonymous pairwise votes.
1,475.829
95% confidence interval: ±8.469The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
Higher measured value1,484.732
95% confidence interval: ±5.864The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
No 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 5,649 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 16,866 anonymous pairwise votes.
1,475.829
95% confidence interval: ±8.469The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
No clear lead1,482.695
95% confidence interval: ±5.961The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
Test 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 7,687 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 20,066 anonymous pairwise votes.
1,450.943
95% confidence interval: ±7.537The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
Higher measured value1,452.218
95% confidence interval: ±5.588The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
No 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 7,687 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 19,090 anonymous pairwise votes.
No clear lead1,450.943
95% confidence interval: ±7.537The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,447.932
95% confidence interval: ±5.678The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
Test 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 350 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 714 anonymous pairwise votes.
1,408.852
95% confidence interval: ±33.025The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
Higher measured value1,458.192
95% confidence interval: ±23.983The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
No 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 350 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 677 anonymous pairwise votes.
1,408.852
95% confidence interval: ±33.025The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
No clear lead1,428.912
95% confidence interval: ±24.225The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
Test 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 406 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 1,105 anonymous pairwise votes.
Higher measured value1,428.58
95% confidence interval: ±30.047The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,417.865
95% confidence interval: ±19.032The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
No 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 406 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 981 anonymous pairwise votes.
1,428.58
95% confidence interval: ±30.047The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
No clear lead1,434.201
95% confidence interval: ±19.922The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
LMArena Text, Style ControlLegal and governmentTest details
Test 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 1,884 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 4,506 anonymous pairwise votes.
1,452.3
95% confidence interval: ±14.019The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
Higher measured value1,473.908
95% confidence interval: ±9.549The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
LMArena Text, Style ControlLegal and governmentTest details
No 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 1,884 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 4,328 anonymous pairwise votes.
1,452.3
95% confidence interval: ±14.019The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
No clear lead1,473.362
95% confidence interval: ±9.718The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
LMArena Text, Style ControlLife, physical and social scienceTest details
Test 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 3,518 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 9,173 anonymous pairwise votes.
1,471.756
95% confidence interval: ±10.373The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
Higher measured value1,480.892
95% confidence interval: ±7.148The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
LMArena Text, Style ControlLife, physical and social scienceTest details
No 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 3,518 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 8,814 anonymous pairwise votes.
1,471.756
95% confidence interval: ±10.373The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
No clear lead1,479.862
95% confidence interval: ±7.218The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
Test 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 10,166 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 25,983 anonymous pairwise votes.
Higher measured value1,471.292
95% confidence interval: ±6.966The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,471.226
95% confidence interval: ±5.337The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
No 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 10,166 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 24,788 anonymous pairwise votes.
No clear lead1,471.292
95% confidence interval: ±6.966The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,464.401
95% confidence interval: ±5.421The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
Test 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 1,019 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 3,058 anonymous pairwise votes.
1,442.604
95% confidence interval: ±18.439The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
Higher measured value1,444.744
95% confidence interval: ±11.133The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
No 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 1,019 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 2,708 anonymous pairwise votes.
1,442.604
95% confidence interval: ±18.439The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
No clear lead1,465.899
95% confidence interval: ±11.797The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
Test 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 1,259 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 3,296 anonymous pairwise votes.
Higher measured value1,483.156
95% confidence interval: ±16.859The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,459.871
95% confidence interval: ±10.89The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
No 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 1,259 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 2,993 anonymous pairwise votes.
No clear lead1,483.156
95% confidence interval: ±16.859The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,467.456
95% confidence interval: ±11.459The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
LMArena Text, Style ControlMedicine and healthcareTest details
Test 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 1,465 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 4,069 anonymous pairwise votes.
1,461.875
95% confidence interval: ±15.828The 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: ±10.122The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
Key to the marks:better valuebetter, but not conclusivetie
Our model tests
Our 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.
Pelican on a bicycle
A community classic for free-form SVG drawing.
Grok 4.6
API access was unavailable for this run. The model was not evaluated.
Reasoning (requested): high
First attempt
Run date
7 Sept 2026
API provider
openrouter
Requested API model
x-ai/grok-4.6
Reasoning setting
high
Pinned endpoint
xAI | x-ai/grok-4.6-20260810
Temperature
Provider default
Token limit including reasoning
8,192 tokens
Test protocol
community-visual-tests-v1
Technical checks
Not run
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.
Reasoning (requested): high
First attempt
Run date
7 Sept 2026
API provider
openrouter
Requested API model
deepseek/deepseek-v4-pro-0813
Reasoning setting
high
Pinned endpoint
DeepSeek | deepseek/deepseek-v4-pro-20260813
Temperature
Provider default
Token limit including reasoning
8,192 tokens
Test protocol
community-visual-tests-v1
Technical checks
Not run
Not yet visually reviewed. A successful recording does not confirm correct physics or full compliance with the prompt.
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 Grok 4.6 and DeepSeek-V4-Pro with other leading models using the same data and benchmark logic.