Head-to-head comparison
GPT-5.6 Sol vs. DeepSeek-V4-Pro
Compare published benchmark results from matching versions and cohorts, API pricing, and technical specifications. Any documented test setup differences remain visible.
219
shared result series
24
distinct benchmarks
2026-08-16
latest retrieval
Change model selection
Specifications and cost
General specifications and standard API pricing
Undisclosed parameter counts and knowledge cutoffs stay marked as unavailable. The table shows direct standard token pricing, including documented context tiers. Batch, fast, priority, regional surcharges, tool calls, and cloud platform pricing are excluded.
| Attribute | GPT-5.6 Sol | DeepSeek-V4-Pro |
|---|---|---|
| Provider | OpenAI | DeepSeek |
| Released | Jun 26, 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 | 1,050,000 tokens | 1,000,000 tokens |
| Knowledge cutoff | February 16, 2026 | Not published |
| Technical sources | Model, Context, Knowledge cutoff | Model, Context |
| API pricing per 1M tokens | ||
| API input | $5 ≤272K / $10 >272K | $0.43 |
| API output | $30 ≤272K / $45 >272K | $0.87 |
| Cache read | $0.5 ≤272K / $1 >272K | $0. |
| Cache write (5 min.) | $6.25 ≤272K / $12.5 >272K | Not listed |
| Cache write (1 hr.) | Not listed | Not listed |
| Price verified | 08/11/2026Source | 07/30/2026Source |
GPT-5.6 SolDeepSeek-V4-Pro
API inputUSD per 1M tokens, base tier
GPT-5.6 Sol$5
DeepSeek-V4-Pro$0.435Lower price
API outputUSD per 1M tokens, base tier
GPT-5.6 Sol$30
DeepSeek-V4-Pro$0.87Lower price
Cache readUSD per 1M tokens, base tier
GPT-5.6 Sol$0.5
DeepSeek-V4-Pro$0.004Lower price
Performance
Shared benchmarks
Only published results with the same benchmark version, task, metric, and comparison cohort are paired. Different reasoning levels, agents, harnesses, or output limits appear directly in the table.
219 shared result series
GPT-5.6 SolDeepSeek-V4-Pro
Arena Text, Style ControlOverall. Higher is better. Test setup not documented as identical
GPT-5.6 Sol1,480.992 arena pointsHigher measured value
DeepSeek-V4-Pro1,458.103 arena points
Code MigrationOverall. Higher is better. No documented setup difference
GPT-5.6 Sol52.916 %Winner
DeepSeek-V4-Pro26.201 %
CorpFin v2Overall. Higher is better. No documented setup difference
GPT-5.6 Sol64.375 %Winner
DeepSeek-V4-Pro61.383 %
CyberBenchOverall. Higher is better. No documented setup difference
GPT-5.6 Sol88.136 %Winner
DeepSeek-V4-Pro68.644 %
EMBOverall. Higher is better. No documented setup difference
GPT-5.6 Sol72.339 %Winner
DeepSeek-V4-Pro51.623 %
Finance Agent (v2)Overall. Higher is better. No documented setup difference
GPT-5.6 Sol53.756 %Winner
DeepSeek-V4-Pro44.083 %
GPQA DiamondOverall. Higher is better. No documented setup difference
GPT-5.6 Sol95.202 %Winner
DeepSeek-V4-Pro89.394 %
Harvey's Legal Agent BenchmarkOverall · Task fully resolved. Higher is better. No documented setup difference
GPT-5.6 Sol2.5 %
DeepSeek-V4-Pro3.75 %Winner
IOIOverall. Higher is better. No documented setup difference
GPT-5.6 Sol86.667 %Winner
DeepSeek-V4-Pro35.833 %
Legal Research BenchOverall · All-pass. Higher is better. No documented setup difference
GPT-5.6 Sol48.077 %Winner
DeepSeek-V4-Pro23.077 %
LegalBenchOverall. Higher is better. No documented setup difference
GPT-5.6 Sol86.965 %Winner
DeepSeek-V4-Pro80.323 %
LiveCodeBench v6Overall. Higher is better. No documented setup difference
GPT-5.6 Sol82.604 %
DeepSeek-V4-Pro87.484 %Winner
| Benchmark and task | GPT-5.6 Sol | DeepSeek-V4-Pro | Comparability | Source |
|---|---|---|---|---|
| Arena Text, Style ControlOverall | 1,480.992 arena pointsHigher measured value 95% confidence interval: ±5.733The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | 1,458.103 arena points 95% confidence interval: ±4.022The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | style-control-v1style-controlled Bradley-Terry ratingHigher is betterCohort: lmarena-text-style-control:2026-08-12:overallTest setup not documented as identicalNo direct winnerGPT-5.6 Sol: Snapshot: gpt-5.6-sol-xhigh, Reasoning: xhigh, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro, Harness: Arena text, style controlGPT-5.6 Sol: Style-controlled Arena rating from 15,450 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 54,288 anonymous pairwise votes. | ArenaType: Official benchmark leaderboardLeaderboard updated: 2026-08-12Import ID: lmarena-eeedf9209373a7f9b324bfb839e7b02fedacd1e5-1611ed64f3f1Retrieved 2026-08-16 |
| Code MigrationOverall | 52.916 %Winner | 26.201 % | 1accuracyHigher is betterCohort: vals-code-migration:1:overallNo documented setup differenceWinner: GPT-5.6 SolGPT-5.6 Sol: No further settings disclosed DeepSeek-V4-Pro: No further settings disclosed | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-13Import ID: vals-code-migration-2026-08-16-107652dc6bfcRetrieved 2026-08-16 |
| CorpFin v2Overall | 64.375 %Winner | 61.383 % | 2accuracyHigher is betterCohort: vals-corp-fin-v2:2:overallNo documented setup differenceWinner: GPT-5.6 SolGPT-5.6 Sol: No further settings disclosed DeepSeek-V4-Pro: No further settings disclosed | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-12Import ID: vals-corp_fin_v2-2026-08-16-45bbf84f57abRetrieved 2026-08-16 |
| CyberBenchOverall | 88.136 %Winner | 68.644 % | 1accuracyHigher is betterCohort: vals-cyberbench:1:overallNo documented setup differenceWinner: GPT-5.6 SolGPT-5.6 Sol: No further settings disclosed DeepSeek-V4-Pro: No further settings disclosed | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-03Import ID: vals-cyber-2026-08-16-702acaee9aa9Retrieved 2026-08-16 |
| EMBOverall | 72.339 %Winner | 51.623 % | 1accuracyHigher is betterCohort: vals-emb:1:overallNo documented setup differenceWinner: GPT-5.6 SolGPT-5.6 Sol: No further settings disclosed DeepSeek-V4-Pro: No further settings disclosed | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-13Import ID: vals-emb-2026-08-16-68d5d1d47a97Retrieved 2026-08-16 |
| Finance Agent (v2)Overall | 53.756 %Winner | 44.083 % | 2accuracyHigher is betterCohort: vals-finance-agent-v2:2:overallNo documented setup differenceWinner: GPT-5.6 SolGPT-5.6 Sol: No further settings disclosed DeepSeek-V4-Pro: No further settings disclosed | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-14Import ID: vals-fabv2-2026-08-16-8853da43085dRetrieved 2026-08-16 |
| GPQA DiamondOverall | 95.202 %Winner | 89.394 % | 1accuracyHigher is betterCohort: vals-gpqa:1:overallNo documented setup differenceWinner: GPT-5.6 SolGPT-5.6 Sol: No further settings disclosed DeepSeek-V4-Pro: No further settings disclosed | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-12Import ID: vals-gpqa-2026-08-16-7f8363eaa2dbRetrieved 2026-08-16 |
| Harvey's Legal Agent BenchmarkOverall · Task fully resolved | 2.5 % | 3.75 %Winner | 1task resolution rateHigher is betterCohort: vals-legal-agent-benchmark:1:overallNo documented setup differenceWinner: DeepSeek-V4-ProGPT-5.6 Sol: No further settings disclosed DeepSeek-V4-Pro: No further settings disclosed | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-14Import ID: vals-hlab-2026-08-16-bae90566df2fRetrieved 2026-08-16 |
| IOIOverall | 86.667 %Winner | 35.833 % | 1accuracyHigher is betterCohort: vals-ioi:1:overallNo documented setup differenceWinner: GPT-5.6 SolGPT-5.6 Sol: No further settings disclosed DeepSeek-V4-Pro: No further settings disclosed | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-09Import ID: vals-ioi-2026-08-16-d9164ef483aaRetrieved 2026-08-16 |
| Legal Research BenchOverall · All-pass | 48.077 %Winner | 23.077 % | 1all-pass rateHigher is betterCohort: vals-legal-research:1:overallNo documented setup differenceWinner: GPT-5.6 SolGPT-5.6 Sol: No further settings disclosed DeepSeek-V4-Pro: No further settings disclosed | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-14Import ID: vals-legal_research-2026-08-16-af296c00ecaaRetrieved 2026-08-16 |
| LegalBenchOverall | 86.965 %Winner | 80.323 % | 1accuracyHigher is betterCohort: vals-legal-bench:1:overallNo documented setup differenceWinner: GPT-5.6 SolGPT-5.6 Sol: No further settings disclosed DeepSeek-V4-Pro: No further settings disclosed | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-12Import ID: vals-legal_bench-2026-08-16-d5b3b1b2b5c8Retrieved 2026-08-16 |
| LiveCodeBench v6Overall | 82.604 % | 87.484 %Winner | 1accuracyHigher is betterCohort: vals-lcb:1:overallNo documented setup differenceWinner: DeepSeek-V4-ProGPT-5.6 Sol: No further settings disclosed DeepSeek-V4-Pro: No further settings disclosed | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-13Import ID: vals-lcb-2026-08-16-17b9c147080dRetrieved 2026-08-16 |
| MedCodeOverall | 43.974 %Winner | 40.455 % | 1accuracyHigher is betterCohort: vals-medcode:1:overallNo documented setup differenceWinner: GPT-5.6 SolGPT-5.6 Sol: No further settings disclosed DeepSeek-V4-Pro: No further settings disclosed | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-12Import ID: vals-medcode-2026-08-16-524f7f474eacRetrieved 2026-08-16 |
| MedScribeOverall | 85.233 %Winner | 75.144 % | 1accuracyHigher is betterCohort: vals-medscribe:1:overallNo documented setup differenceWinner: GPT-5.6 SolGPT-5.6 Sol: No further settings disclosed DeepSeek-V4-Pro: No further settings disclosed | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-12Import ID: vals-medscribe-2026-08-16-c385354a7e62Retrieved 2026-08-16 |
| MMLU-ProOverall | 89.1 %Winner | 87.249 % | 1accuracyHigher is betterCohort: vals-mmlu-pro:1:overallNo documented setup differenceWinner: GPT-5.6 SolGPT-5.6 Sol: No further settings disclosed DeepSeek-V4-Pro: No further settings disclosed | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-12Import ID: vals-mmlu_pro-2026-08-16-e14d5bb3a34cRetrieved 2026-08-16 |
| ProgramBenchOverall | 1.5 %Winner | 0 % | 1accuracyHigher is betterCohort: vals-programbench:1:overallNo documented setup differenceWinner: GPT-5.6 SolGPT-5.6 Sol: No further settings disclosed DeepSeek-V4-Pro: No further settings disclosed | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-13Import ID: vals-programbench-2026-08-16-08d9727662ddRetrieved 2026-08-16 |
| ProofBenchOverall | 83 %Winner | 16 % | 1.1accuracyHigher is betterCohort: vals-proof-bench:1.1:overallNo documented setup differenceWinner: GPT-5.6 SolGPT-5.6 Sol: No further settings disclosed DeepSeek-V4-Pro: No further settings disclosed | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-14Import ID: vals-proof_bench-2026-08-16-14232db24de1Retrieved 2026-08-16 |
| Public Benefits Bench v1.1Overall | 66.509 %Winner | 62.923 % | 1.1accuracyHigher is betterCohort: vals-snap:1.1:overallNo documented setup differenceWinner: GPT-5.6 SolGPT-5.6 Sol: No further settings disclosed DeepSeek-V4-Pro: No further settings disclosed | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-12Import ID: vals-public-benefits-bench-2026-08-16-7ce68123c502Retrieved 2026-08-16 |
| SkillsBenchOverall | 54.1 %Winner | 51.274 % | 1accuracyHigher is betterCohort: vals-skillsbench:1:overallNo documented setup differenceWinner: GPT-5.6 SolGPT-5.6 Sol: Harness: OpenHands DeepSeek-V4-Pro: Harness: OpenHands | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-13Import ID: vals-skillsbench-2026-08-16-4917a67268d2Retrieved 2026-08-16 |
| SWE-bench VerifiedOverall | 96.2 %Winner | 77.4 % | 1accuracyHigher is betterCohort: vals-swebench:1:overallNo documented setup differenceWinner: GPT-5.6 SolGPT-5.6 Sol: No further settings disclosed DeepSeek-V4-Pro: No further settings disclosed | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-12Import ID: vals-swebench-2026-08-16-c133071b8167Retrieved 2026-08-16 |
| TaxEval v2Overall | 74.775 %Winner | 72.077 % | 2accuracyHigher is betterCohort: vals-tax-eval-v2:2:overallNo documented setup differenceWinner: GPT-5.6 SolGPT-5.6 Sol: No further settings disclosed DeepSeek-V4-Pro: No further settings disclosed | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-12Import ID: vals-tax_eval_v2-2026-08-16-5bbd495dac53Retrieved 2026-08-16 |
| Terminal-Bench 2.1Overall | 85.768 %Winner | 50.187 % | 2.1accuracyHigher is betterCohort: vals-terminal-bench-2-1:2.1:overallNo documented setup differenceWinner: GPT-5.6 SolGPT-5.6 Sol: No further settings disclosed DeepSeek-V4-Pro: No further settings disclosed | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-13Import ID: vals-terminal-bench-2-1-2026-08-16-b3f38c18bea4Retrieved 2026-08-16 |
| Vals IndexOverall | 63.709 %Winner | 42.888 % | 2weighted index scoreHigher is betterCohort: vals-vals-index:2:overallNo documented setup differenceWinner: GPT-5.6 SolGPT-5.6 Sol: No further settings disclosed DeepSeek-V4-Pro: No further settings disclosed | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-14Import ID: vals-vals_index-2026-08-16-ab641d89d6edRetrieved 2026-08-16 |
| Vibe Code Bench v1.1Overall | 80.495 %Winner | 49.931 % | 1.1accuracyHigher is betterCohort: vals-vibe-code-bench:1.1:overallNo documented setup differenceWinner: GPT-5.6 SolGPT-5.6 Sol: Harness: OpenHands DeepSeek-V4-Pro: Harness: OpenHands | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-13Import ID: vals-vibe-code-2026-08-16-a73c9770bcc4Retrieved 2026-08-16 |
| Arena Text, Style ControlOverall | 1,480.992 arena pointsHigher measured value 95% confidence interval: ±5.733The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | 1,455.61 arena points 95% confidence interval: ±4.089The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | style-control-v1style-controlled Bradley-Terry ratingHigher is betterCohort: lmarena-text-style-control:2026-08-12:overallTest setup not documented as identicalNo direct winnerGPT-5.6 Sol: Snapshot: gpt-5.6-sol-xhigh, Reasoning: xhigh, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro-high-preview, Reasoning: high, Harness: Arena text, style controlGPT-5.6 Sol: Style-controlled Arena rating from 15,450 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 51,670 anonymous pairwise votes. | ArenaType: Official benchmark leaderboardLeaderboard updated: 2026-08-12Import ID: lmarena-eeedf9209373a7f9b324bfb839e7b02fedacd1e5-1611ed64f3f1Retrieved 2026-08-16 |
| Code MigrationCLIReasoning: max | 47.221 %Higher measured value | 15.47 % | 1accuracyHigher is betterCohort: vals-code-migration:1:cliTest setup not documented as identicalNo direct winnerGPT-5.6 Sol: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4-Pro: Reasoning: max, Output limit: 384,000 tokens | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-09Import ID: vals-code-migration-2026-08-11-3a2d5ce718a3Retrieved 2026-08-11 |
| CorpFin v2Exact PagesReasoning: max | 61.538 %Winner | 59.207 % | 2accuracyHigher is betterCohort: vals-corp-fin-v2:2:exact_pagesNo documented setup differenceWinner: GPT-5.6 SolGPT-5.6 Sol: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4-Pro: Reasoning: max, Output limit: 128,000 tokens | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-06Import ID: vals-corp_fin_v2-2026-08-11-d9e58d191cd9Retrieved 2026-08-11 |
| CyberBenchPatch | 83.051 %Winner | 79.661 % | 1accuracyHigher is betterCohort: vals-cyberbench:1:patchNo documented setup differenceWinner: GPT-5.6 SolGPT-5.6 Sol: No further settings disclosed DeepSeek-V4-Pro: No further settings disclosed | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-03Import ID: vals-cyber-2026-08-11-1f20498a3dc1Retrieved 2026-08-11 |
| EMBCompsReasoning: max | 71.309 %Higher measured value | 57.434 % | 1accuracyHigher is betterCohort: vals-emb:1:compsTest setup not documented as identicalNo direct winnerGPT-5.6 Sol: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4-Pro: Reasoning: max, Output limit: 384,000 tokens | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-06Import ID: vals-emb-2026-08-11-e3c855e236e4Retrieved 2026-08-11 |
| Finance Agent (v2)AdjustmentsReasoning: max | 49.233 %Higher measured value | 41.586 % | 2accuracyHigher is betterCohort: vals-finance-agent-v2:2:adjustmentsTest setup not documented as identicalNo direct winnerGPT-5.6 Sol: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4-Pro: Reasoning: max, Output limit: 384,000 tokens | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-06Import ID: vals-fabv2-2026-08-11-3e5c93a1857cRetrieved 2026-08-11 |
| GPQA DiamondFew-Shot CoTReasoning: max | 94.949 %Higher measured value | 90.404 % | 1accuracyHigher is betterCohort: vals-gpqa:1:diamond_few_shot_cotTest setup not documented as identicalNo direct winnerGPT-5.6 Sol: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4-Pro: Reasoning: max, Output limit: 256,000 tokens | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-09Import ID: vals-gpqa-2026-08-11-d3325290a404Retrieved 2026-08-11 |
| Harvey's Legal Agent BenchmarkAntitrust Competition · Criteria passedReasoning: max | 78.623 % | 83.696 %Higher measured value | 1criteria pass rateHigher is betterCohort: vals-legal-agent-benchmark:1:criteria_pass_rate__antitrust_competitionTest setup not documented as identicalNo direct winnerGPT-5.6 Sol: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4-Pro: Reasoning: max, Output limit: 384,000 tokens | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-07Import ID: vals-hlab-2026-08-11-ea11b6e53414Retrieved 2026-08-11 |
| IOIIOI 2024Reasoning: max | 85 %Higher measured value | 31.333 % | 1accuracyHigher is betterCohort: vals-ioi:1:ioi2024Test setup not documented as identicalNo direct winnerGPT-5.6 Sol: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4-Pro: Reasoning: max, Output limit: 256,000 tokens | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-09Import ID: vals-ioi-2026-08-11-2dfe8b4b8b7bRetrieved 2026-08-11 |
| Legal Research BenchAdministrative / Regulatory · All-passReasoning: max | 60.606 %Higher measured value | 28.788 % | 1all-pass rateHigher is betterCohort: vals-legal-research:1:administrative_regulatoryTest setup not documented as identicalNo direct winnerGPT-5.6 Sol: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4-Pro: Reasoning: max, Output limit: 384,000 tokens | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-06Import ID: vals-legal_research-2026-08-11-326cf7aaca58Retrieved 2026-08-11 |
| LegalBenchConclusion TasksReasoning: max | 90.492 %Higher measured value | 89.117 % | 1accuracyHigher is betterCohort: vals-legal-bench:1:conclusion_tasksTest setup not documented as identicalNo direct winnerGPT-5.6 Sol: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4-Pro: Reasoning: max, Output limit: 256,000 tokens | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-09Import ID: vals-legal_bench-2026-08-11-836e5e44618eRetrieved 2026-08-11 |
| LiveCodeBench v6EasyReasoning: max | 98.137 % | 98.758 %Winner | 1accuracyHigher is betterCohort: vals-lcb:1:easyNo documented setup differenceWinner: DeepSeek-V4-ProGPT-5.6 Sol: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4-Pro: Reasoning: max, Output limit: 128,000 tokens | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-09Import ID: vals-lcb-2026-08-11-0342bcc364e4Retrieved 2026-08-11 |
| MMLU-ProBiologyReasoning: max | 94.561 %Higher measured value | 92.887 % | 1accuracyHigher is betterCohort: vals-mmlu-pro:1:biologyTest setup not documented as identicalNo direct winnerGPT-5.6 Sol: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4-Pro: Reasoning: max, Output limit: 256,000 tokens | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-09Import ID: vals-mmlu_pro-2026-08-11-b8e53828d746Retrieved 2026-08-11 |
| ProgramBenchAlmost ResolvedReasoning: max | 23 %Higher measured value | 0 % | 1accuracyHigher is betterCohort: vals-programbench:1:almostTest setup not documented as identicalNo direct winnerGPT-5.6 Sol: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4-Pro: Reasoning: max, Output limit: 384,000 tokens | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-09Import ID: vals-programbench-2026-08-11-b6f8959fbdfeRetrieved 2026-08-11 |
| SkillsBenchNo SkillsReasoning: max | 43.312 %Higher measured value | 27.226 % | 1accuracyHigher is betterCohort: vals-skillsbench:1:no_skillsTest setup not documented as identicalNo direct winnerGPT-5.6 Sol: Reasoning: max, Harness: OpenHands, Output limit: 128,000 tokens DeepSeek-V4-Pro: Reasoning: max, Harness: OpenHands, Output limit: 384,000 tokens | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-07Import ID: vals-skillsbench-2026-08-11-351bec9d0e01Retrieved 2026-08-11 |
| SWE-bench Verified<15 min fixReasoning: max | 97.423 %Higher measured value | 87.629 % | 1accuracyHigher is betterCohort: vals-swebench:1:<15 min fixTest setup not documented as identicalNo direct winnerGPT-5.6 Sol: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4-Pro: Reasoning: max, Output limit: 256,000 tokens | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-08Import ID: vals-swebench-2026-08-11-a6212cd857d8Retrieved 2026-08-11 |
| TaxEval v2CorrectnessReasoning: max | 65.004 %Winner | 60.589 % | 2accuracyHigher is betterCohort: vals-tax-eval-v2:2:correctnessNo documented setup differenceWinner: GPT-5.6 SolGPT-5.6 Sol: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4-Pro: Reasoning: max, Output limit: 128,000 tokens | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-09Import ID: vals-tax_eval_v2-2026-08-11-d05a40bf446fRetrieved 2026-08-11 |
| Terminal-Bench 2.1Easy | 91.667 %Higher measured value | 66.667 % | 2.1accuracyHigher is betterCohort: vals-terminal-bench-2-1:2.1:easyTest setup not documented as identicalNo direct winnerGPT-5.6 Sol: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4-Pro: Output limit: 128,000 tokens | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-06Import ID: vals-terminal-bench-2-1-2026-08-11-8f0ebd72a609Retrieved 2026-08-11 |
| Vals IndexCorpFin v2Reasoning: max | 66.434 %Winner | 61.888 % | 1.2benchmark scoreHigher is betterCohort: vals-vals-index:1.2:corp_fin_v2No documented setup differenceWinner: GPT-5.6 SolGPT-5.6 Sol: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4-Pro: Reasoning: max, Output limit: 128,000 tokens | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-09Import ID: vals-vals_index-2026-08-11-e79d08d59249Retrieved 2026-08-11 |
| Arena Text, Style ControlBusiness, management and finance | 1,482.723 arena pointsHigher measured value 95% confidence interval: ±11.356The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | 1,460.638 arena points 95% confidence interval: ±6.951The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | style-control-v1style-controlled Bradley-Terry ratingHigher is betterCohort: lmarena-text-style-control:2026-08-12:industry_business_and_management_and_financial_operationsTest setup not documented as identicalNo direct winnerGPT-5.6 Sol: Snapshot: gpt-5.6-sol-xhigh, Reasoning: xhigh, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro, Harness: Arena text, style controlGPT-5.6 Sol: Style-controlled Arena rating from 3,027 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 10,715 anonymous pairwise votes. | ArenaType: Official benchmark leaderboardLeaderboard updated: 2026-08-12Import ID: lmarena-eeedf9209373a7f9b324bfb839e7b02fedacd1e5-1611ed64f3f1Retrieved 2026-08-16 |
| Code MigrationCOBOLReasoning: max | 70 %Higher measured value | 58.393 % | 1accuracyHigher is betterCohort: vals-code-migration:1:cobolTest setup not documented as identicalNo direct winnerGPT-5.6 Sol: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4-Pro: Reasoning: max, Output limit: 384,000 tokens | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-09Import ID: vals-code-migration-2026-08-11-3a2d5ce718a3Retrieved 2026-08-11 |
| CorpFin v2Max Fitting ContextReasoning: max | 65.152 %Winner | 63.054 % | 2accuracyHigher is betterCohort: vals-corp-fin-v2:2:max_fitting_contextNo documented setup differenceWinner: GPT-5.6 SolGPT-5.6 Sol: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4-Pro: Reasoning: max, Output limit: 128,000 tokens | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-06Import ID: vals-corp_fin_v2-2026-08-11-d9e58d191cd9Retrieved 2026-08-11 |
| CyberBenchPoC | 93.22 %Winner | 57.627 % | 1accuracyHigher is betterCohort: vals-cyberbench:1:pocNo documented setup differenceWinner: GPT-5.6 SolGPT-5.6 Sol: No further settings disclosed DeepSeek-V4-Pro: No further settings disclosed | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-03Import ID: vals-cyber-2026-08-11-1f20498a3dc1Retrieved 2026-08-11 |
| EMBDataroomReasoning: max | 85.243 %Higher measured value | 62.572 % | 1accuracyHigher is betterCohort: vals-emb:1:dataroomTest setup not documented as identicalNo direct winnerGPT-5.6 Sol: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4-Pro: Reasoning: max, Output limit: 384,000 tokens | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-06Import ID: vals-emb-2026-08-11-e3c855e236e4Retrieved 2026-08-11 |
| Finance Agent (v2)All-PassReasoning: max | 41.362 %Higher measured value | 31.448 % | 2accuracyHigher is betterCohort: vals-finance-agent-v2:2:all_passTest setup not documented as identicalNo direct winnerGPT-5.6 Sol: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4-Pro: Reasoning: max, Output limit: 384,000 tokens | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-06Import ID: vals-fabv2-2026-08-11-3e5c93a1857cRetrieved 2026-08-11 |
| GPQA DiamondZero-Shot CoTReasoning: max | 95.455 %Higher measured value | 88.384 % | 1accuracyHigher is betterCohort: vals-gpqa:1:diamond_zero_shot_cotTest setup not documented as identicalNo direct winnerGPT-5.6 Sol: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4-Pro: Reasoning: max, Output limit: 256,000 tokens | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-09Import ID: vals-gpqa-2026-08-11-d3325290a404Retrieved 2026-08-11 |
| Harvey's Legal Agent BenchmarkAntitrust Competition · Task fully resolvedReasoning: max | 0 %No direct winner | 0 %No direct winner | 1task resolution rateHigher is betterCohort: vals-legal-agent-benchmark:1:antitrust_competitionTest setup not documented as identicalNo direct winnerGPT-5.6 Sol: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4-Pro: Reasoning: max, Output limit: 384,000 tokens | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-07Import ID: vals-hlab-2026-08-11-ea11b6e53414Retrieved 2026-08-11 |
| IOIIOI 2025Reasoning: max | 88.333 %Higher measured value | 40.333 % | 1accuracyHigher is betterCohort: vals-ioi:1:ioi2025Test setup not documented as identicalNo direct winnerGPT-5.6 Sol: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4-Pro: Reasoning: max, Output limit: 256,000 tokens | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-09Import ID: vals-ioi-2026-08-11-2dfe8b4b8b7bRetrieved 2026-08-11 |
| Legal Research BenchAdministrative / Regulatory · Weighted partial creditReasoning: max | 90.69 %Higher measured value | 83.372 % | 1weighted partial-credit scoreHigher is betterCohort: vals-legal-research:1:weighted_pass__administrative_regulatoryTest setup not documented as identicalNo direct winnerGPT-5.6 Sol: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4-Pro: Reasoning: max, Output limit: 384,000 tokens | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-06Import ID: vals-legal_research-2026-08-11-326cf7aaca58Retrieved 2026-08-11 |
| LegalBenchInterpretation TasksReasoning: max | 84.398 %Higher measured value | 75.169 % | 1accuracyHigher is betterCohort: vals-legal-bench:1:interpretation_tasksTest setup not documented as identicalNo direct winnerGPT-5.6 Sol: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4-Pro: Reasoning: max, Output limit: 256,000 tokens | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-09Import ID: vals-legal_bench-2026-08-11-836e5e44618eRetrieved 2026-08-11 |
| LiveCodeBench v6HardReasoning: max | 67.429 % | 72.571 %Winner | 1accuracyHigher is betterCohort: vals-lcb:1:hardNo documented setup differenceWinner: DeepSeek-V4-ProGPT-5.6 Sol: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4-Pro: Reasoning: max, Output limit: 128,000 tokens | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-09Import ID: vals-lcb-2026-08-11-0342bcc364e4Retrieved 2026-08-11 |
| MMLU-ProBusinessReasoning: max | 93.283 %Higher measured value | 91.888 % | 1accuracyHigher is betterCohort: vals-mmlu-pro:1:businessTest setup not documented as identicalNo direct winnerGPT-5.6 Sol: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4-Pro: Reasoning: max, Output limit: 256,000 tokens | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-09Import ID: vals-mmlu_pro-2026-08-11-b8e53828d746Retrieved 2026-08-11 |
| ProgramBenchFully ResolvedReasoning: max | 1.5 %Higher measured value | 0 % | 1accuracyHigher is betterCohort: vals-programbench:1:strictTest setup not documented as identicalNo direct winnerGPT-5.6 Sol: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4-Pro: Reasoning: max, Output limit: 384,000 tokens | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-09Import ID: vals-programbench-2026-08-11-b6f8959fbdfeRetrieved 2026-08-11 |
| SkillsBenchWith SkillsReasoning: max | 54.1 %Higher measured value | 51.274 % | 1accuracyHigher is betterCohort: vals-skillsbench:1:with_skillsTest setup not documented as identicalNo direct winnerGPT-5.6 Sol: Reasoning: max, Harness: OpenHands, Output limit: 128,000 tokens DeepSeek-V4-Pro: Reasoning: max, Harness: OpenHands, Output limit: 384,000 tokens | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-07Import ID: vals-skillsbench-2026-08-11-351bec9d0e01Retrieved 2026-08-11 |
| SWE-bench Verified>4 hoursReasoning: max | 100 %Higher measured value | 0 % | 1accuracyHigher is betterCohort: vals-swebench:1:>4 hoursTest setup not documented as identicalNo direct winnerGPT-5.6 Sol: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4-Pro: Reasoning: max, Output limit: 256,000 tokens | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-08Import ID: vals-swebench-2026-08-11-a6212cd857d8Retrieved 2026-08-11 |
| TaxEval v2Stepwise ReasoningReasoning: max | 84.546 %Winner | 83.565 % | 2accuracyHigher is betterCohort: vals-tax-eval-v2:2:stepwise_reasoningNo documented setup differenceWinner: GPT-5.6 SolGPT-5.6 Sol: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4-Pro: Reasoning: max, Output limit: 128,000 tokens | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-09Import ID: vals-tax_eval_v2-2026-08-11-d05a40bf446fRetrieved 2026-08-11 |
| Terminal-Bench 2.1Hard | 84.444 %Higher measured value | 31.111 % | 2.1accuracyHigher is betterCohort: vals-terminal-bench-2-1:2.1:hardTest setup not documented as identicalNo direct winnerGPT-5.6 Sol: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4-Pro: Output limit: 128,000 tokens | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-06Import ID: vals-terminal-bench-2-1-2026-08-11-8f0ebd72a609Retrieved 2026-08-11 |
| Vals IndexFinance Agent v2Reasoning: max | 54.955 %Higher measured value | 44.081 % | 1.2benchmark scoreHigher is betterCohort: vals-vals-index:1.2:finance_agentTest setup not documented as identicalNo direct winnerGPT-5.6 Sol: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4-Pro: Reasoning: max, Output limit: 384,000 tokens | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-09Import ID: vals-vals_index-2026-08-11-e79d08d59249Retrieved 2026-08-11 |
| Arena Text, Style ControlBusiness, management and finance | 1,482.723 arena pointsHigher measured value 95% confidence interval: ±11.356The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | 1,447.754 arena points 95% confidence interval: ±7.037The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result. | style-control-v1style-controlled Bradley-Terry ratingHigher is betterCohort: lmarena-text-style-control:2026-08-12:industry_business_and_management_and_financial_operationsTest setup not documented as identicalNo direct winnerGPT-5.6 Sol: Snapshot: gpt-5.6-sol-xhigh, Reasoning: xhigh, Harness: Arena text, style control DeepSeek-V4-Pro: Snapshot: deepseek-v4-pro-high-preview, Reasoning: high, Harness: Arena text, style controlGPT-5.6 Sol: Style-controlled Arena rating from 3,027 anonymous pairwise votes.DeepSeek-V4-Pro: Style-controlled Arena rating from 10,363 anonymous pairwise votes. | ArenaType: Official benchmark leaderboardLeaderboard updated: 2026-08-12Import ID: lmarena-eeedf9209373a7f9b324bfb839e7b02fedacd1e5-1611ed64f3f1Retrieved 2026-08-16 |
| Code MigrationCode QualityReasoning: max | 3.847 / 5Higher measured value | 3.401 / 5 | 1mean reviewer code-quality ratingHigher is betterCohort: vals-code-migration:1:code_qualityTest setup not documented as identicalNo direct winnerGPT-5.6 Sol: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4-Pro: Reasoning: max, Output limit: 384,000 tokens | Vals AIType: Independent evaluationLeaderboard updated: 2026-08-09Import ID: vals-code-migration-2026-08-11-3a2d5ce718a3Retrieved 2026-08-11 |
The table shows 64 diversified result series from 219 shared results. This prevents one benchmark with many subtasks from crowding out every other test.
How to read this comparison
Winning one benchmark is not an overall verdict
Your workload, budget, and required context length matter most. A coding benchmark says little about visual understanding or agent performance.
Prices use official API rates per 1M tokens. Web subscriptions and cloud platform surcharges can differ.
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.