AIEvals

Models

GPT-6 Sol

OpenAI’s GPT-6 Sol release, retained alongside its GPT-6.1 successor for comparison.

OpenAIReleased 22 September 2026Context 1.05M tokensOutput 128k tokensPrivate weights$2.00 in / $10.00 out per Mtok

Family GPT-641 benchmarks with a published result1 independent first place among the 46 models this site listsRead 17 October 2026

Where it stands

Rank compares tracked models with a score. Published shows the source’s rank and field size, when available.

BenchmarkScoreRankPublishedLeaderBehind
Vals IndexVals AI57.54%11of 3311of 41Gemini 4 Argon11.36 pts
Artificial Analysis Intelligence IndexArtificial Analysis489of 42no list readClaude Opus 5.510 pts
Terminal-Bench 4.0Artificial Analysis43.94%8of 40no list readClaude Sonnet 5.519.70 pts
Vibe Code Bench v1.1Vals AI87.82%9of 389of 106Claude Sonnet 5.54.57 pts
Terminal-Bench Science 0.1Artificial Analysis30.00%6of 24no list readGPT-6 Astra33.33 pts
ProgramBenchVals AI2.00%6of 296of 54Claude Opus 5.516.50 pts
ITBench-AAArtificial Analysis49.44%6of 18no list readGPT-5.6 Sol6.77 pts
MLCR-AAArtificial Analysis16.11%28of 37no list readClaude Sonnet 5.558.89 pts
ARC-AGI-3ARC Prize · Standard4.62%5of 11no list readGPT-6 Astra58.09 pts
Vals RSI IndexVals AI28.11%5of 205of 23Claude Opus 5.59.20 pts
AutomationBench-AAArtificial Analysis61.63%12of 40no list readGemini 4 Argon15.88 pts
SciCodeArtificial Analysis57.64%12of 40no list readClaude Opus 5.59.26 pts
GDP.pdfArtificial Analysis25.20%8of 40no list readGPT-6.1 Sol5.80 pts
CritPtArtificial Analysis30.86%6of 42no list readGPT-5.6 Sol1.43 pts
AA-OmniscienceArtificial Analysis27.1216of 42no list readClaude Opus 5.519.3
AA-LCR v1.1Artificial Analysis83.67%9of 42no list readKimi K35.00 pts
Terminal-Bench 2.1Vals AI83.15%6of 356of 75Claude Opus 5.54.49 pts
ProofBench v1.1Vals AI83.00%10of 3412of 46Claude Sonnet 5.517.00 pts
Harvey's Legal Agent BenchmarkVals AI1.67%33of 3744of 73Muse Spark 1.223.75 pts
MysteryMechanismVals AI30.18%12of 2112of 21GPT-6 Astra22.97 pts
BioMysteryBenchVals AI74.81%7of 197of 21Claude Sonnet 5.56.30 pts
Code MigrationVals AI57.20%7of 377of 71Claude Sonnet 5.512.63 pts
CyberBench v1.1Vals AI77.98%1of 311of 43GPT-6 Sol—
EMBVals AI71.53%10of 3710of 69Claude Fable 5.15.14 pts
Finance Agent (v2)Vals AI49.05%32of 3741of 73Gemini 4 Argon16.35 pts
IOIVals AI82.61%10of 3110of 38Gemini 4 Argon17.39 pts
Legal Research BenchVals AI28.85%34of 3742of 72Muse Spark 1.3 Max26.44 pts
MedCodeVals AI47.07%19of 3534of 104Claude Opus 516.50 pts
MedScribeVals AI82.03%32of 3650of 106Claude Opus 5.59.40 pts
Public Benefits Bench v1.1Vals AI56.63%28of 2934of 45Claude Opus 520.30 pts
SAGEVals AI44.79%24of 3243of 90Claude Opus 4.711.31 pts
Tax Agent BenchVals AI53.05%33of 3740of 64Claude Fable 5.124.59 pts
GDPval-AA v2.1Artificial Analysis1,505.2225of 42no list readClaude Opus 5.5340.95
AA-Briefcase v1.1Artificial Analysis1,478.9720of 41no list readClaude Opus 5.5342.88
Humanity's Last ExamArtificial Analysis47.91%13of 42no list readClaude Opus 5.513.44 pts

Nearby on Vals Index

The two models above and below on the Vals Index (Vals AI), on every column where this model has a published number.

ModelVals Index%Artificial Analysis Intelligence IndexindexTerminal-Bench 4.0%Vibe Code Bench v1.1%Terminal-Bench Science 0.1%ProgramBench%
Muse Spark 1.3 MaxMeta58.16%4833.33%85.86%10.95%–
GPT-5.6 SolOpenAI58.01%4739.90%80.50%22.38%1.50%
GPT-6 SolOpenAI57.54%4843.94%87.82%30.00%2.00%
MiMo V2.6 ProXiaomi55.20%4634.85%85.22%5.71%0.50%
Claude Opus 4.8Anthropic55.10%4221.72%82.72%–1.00%
Bold marks a figure better than this model’s on the same column, whichever model it belongs to. Ranks cover the models shown from the stated source.

Cost and speed, as published

Additional measurements from Artificial Analysis’ own row. Read 30 September 2026.

As publishedCost per taskOutput speed
GPT-6 Sol (max)$1.0576 tok/s

Results by benchmark

Independent and vendor results retain their own harnesses and effort settings.

BenchmarkIndependent runAs reported by a publisher
ITBench-AAIT operations49.44%Artificial Analysisnot published
MLCR-AAMedical reasoning16.11%Artificial Analysisnot published
ARC-AGI-3Interactive reasoning4.62%ARC Prize · Standard
23.03%ARC Prize · Provider adapter
not published
Vals RSI IndexAutonomous research28.11%Vals AInot published
AutomationBench-AAWorkflow automation61.63%Artificial Analysisnot published
SciCodeScientific coding57.64%Artificial Analysisnot published
GDP.pdfDocument reasoning25.20%Artificial Analysisnot published
CritPtScientific reasoning30.86%Artificial Analysisnot published
AA-OmniscienceFactual knowledge27.12Artificial Analysisnot published
AA-LCR v1.1Long-context reasoning83.67%Artificial Analysisnot published
Terminal-Bench 4.0Agentic coding43.94%Artificial Analysis
44.44%Vals AI
not published
Terminal-Bench 2.1Agentic coding83.15%Vals AInot published
MMMU ProMultimodal reasoning
83.29%Artificial Analysis
not published
Vibe Code Bench v1.1Product coding87.82%Vals AInot published
Terminal-Bench Science 0.1Agentic science30.00%Artificial Analysis
30.00%Vals AI
27.6%Terminal-Bench Science 0.1OpenAI · 29 September 2026Rank 4 of 4 in this publisher table
ProgramBenchAgentic coding2.00%Vals AInot published
ProofBench v1.1Mathematics83.00%Vals AInot published
Harvey's Legal Agent BenchmarkProfessional knowledge1.67%Vals AInot published
MysteryMechanismScientific reasoning30.18%Vals AInot published
Vals IndexComposite57.54%Vals AInot published
BioMysteryBenchAgentic science74.81%Vals AInot published
Code MigrationAgentic coding57.20%Vals AInot published
CyberBench v1.1Security77.98%Vals AInot published
EMBFinancial modelling71.53%Vals AInot published
Finance Agent (v2)Financial analysis49.05%Vals AInot published
IOICompetitive programming82.61%Vals AInot published
Legal Research BenchProfessional knowledge28.85%Vals AInot published
MedCodeHealthcare administration47.07%Vals AInot published
MedScribeHealthcare administration82.03%Vals AInot published
Public Benefits Bench v1.1Public services56.63%Vals AInot published
SAGEEducation44.79%Vals AInot published
Tax Agent BenchProfessional knowledge53.05%Vals AInot published
OSWorld 2.0Computer useno published result
64.4%OSWorld 2.0 (offline set, partial reward)OpenAI · 29 September 2026Rank 3 of 3 in this publisher table
AutomationBenchBusiness workflowsno published result
32.0%AutomationBench 1.0.6OpenAI · 29 September 2026Rank 4 of 5 in this publisher table
DeepSWE v1.1Software engineeringno published result
68.8%DeepSWE v1.1OpenAI · 29 September 2026Rank 3 of 3 in this publisher table
FrontierCode v1.1Agentic codingno published result
49.3%FrontierCode 1.1 (Main)Anthropic · 28 September 2026Rank 2 of 4 in this publisher table
GDPval-AA v2.1Knowledge work1,505.22Artificial Analysis
1,487GDPval-AA v2.1Anthropic · 28 September 2026Rank 3 of 4 in this publisher table
AA-Briefcase v1.1Knowledge work1,478.97Artificial Analysis
1,483AA-Briefcase v1.1Anthropic · 28 September 2026Rank 3 of 4 in this publisher table
ChartographyVisual chart recognitionno published result
53.6%Chartography (no tools)Anthropic · 28 September 2026Rank 3 of 4 in this publisher table
Humanity's Last ExamMultidisciplinary reasoning47.91%Artificial Analysisnot published
Artificial Analysis Intelligence IndexComposite48Artificial Analysisnot published
Independent column: the suite named under each figure, one harness per suite. Publisher column: each lab’s own table, reproduced on the benchmark’s page. Publisher ranks compare reported scores in the same table row only; ties share a rank and unreported cells are excluded.

Sources

Read on 17 October 2026.