AIEvals

Models

Claude Fable 5.1

Anthropic’s research-oriented Fable release, with published coding, scientific and knowledge-work evaluations.

AnthropicReleased 1 September 2026Context 1M tokensOutput 128k tokensPrivate weights$10.00 in / $50.00 out per Mtok

Family Claude Fable 566 benchmarks with a published result7 independent first places 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 AI65.83%4of 334of 41Gemini 4 Argon3.07 pts
Artificial Analysis Intelligence IndexArtificial Analysis533of 42no list readClaude Opus 5.55 pts
Terminal-Bench 4.0Artificial Analysis52.02%6of 40no list readClaude Sonnet 5.511.62 pts
LiveCodeBenchVals AI90.52%1of 211of 143Claude Fable 5.1—
GPQA DiamondVals AI93.43%7of 237of 138Gemini 3.1 Pro2.02 pts
MMLU ProVals AI92.38%1of 241of 138Claude Fable 5.1—
MMMU ProVals AI90.64%1of 201of 93Claude Fable 5.1—
Vibe Code Bench v1.1Vals AI90.26%5of 385of 106Claude Sonnet 5.52.13 pts
Terminal-Bench Science 0.1Artificial Analysis43.33%5of 24no list readGPT-6 Astra20.00 pts
ProgramBenchVals AI7.00%2of 292of 54Claude Opus 5.511.50 pts
τ³-BankingArtificial Analysis47.22%6of 30no list readQwen3.8 Max4.12 pts
ITBench-AAArtificial Analysis49.53%5of 18no list readGPT-5.6 Sol6.68 pts
AA-AnalystAgentArtificial Analysis57.50%2of 18no list readGemini 3.7 Flash2.50 pts
MLCR-AAArtificial Analysis71.11%2of 37no list readClaude Sonnet 5.53.89 pts
SWE-Bench Pro V2 FullScale AI / SEAL99.10%2of 10no list readClaude Opus 50.30 pts
SWE-Bench Pro V2 HARDScale AI / SEAL92.20%2of 10no list readClaude Opus 55.80 pts
Vals RSI IndexVals AI36.09%2of 202of 23Claude Opus 5.51.22 pts
Vibe Code Bench 1–100Vals AI28.00%3of 203of 20Claude Opus 5.52.36 pts
CUA-benchVals AI13.17%3of 83of 8GPT-6 Astra6.00 pts
Harvey LAB-AAArtificial Analysis93.02%7of 20no list readMuse Spark 1.3 Max2.46 pts
AutomationBench-AAArtificial Analysis59.38%17of 40no list readGemini 4 Argon18.13 pts
SciCodeArtificial Analysis63.08%2of 40no list readClaude Opus 5.53.82 pts
GDP.pdfArtificial Analysis26.20%5of 40no list readGPT-6.1 Sol4.80 pts
CritPtArtificial Analysis29.71%8of 42no list readGPT-5.6 Sol2.58 pts
AA-OmniscienceArtificial Analysis43.452of 42no list readClaude Opus 5.52.97
AA-LCR v1.1Artificial Analysis85.33%4of 42no list readKimi K33.34 pts
Terminal-Bench 2.1Vals AI85.02%4of 354of 75Claude Opus 5.52.62 pts
ProofBench v1.1Vals AI100.00%1of 341of 46Claude Sonnet 5.5—
LegalBenchVals AI88.51%2of 272of 149Claude Fable 50.05 pts
Harvey's Legal Agent BenchmarkVals AI6.67%18of 3724of 73Muse Spark 1.218.75 pts
SkillsBenchVals AI61.55%5of 206of 35DeepSeek V4.1 Flash8.25 pts
MysteryMechanismVals AI47.75%4of 214of 21GPT-6 Astra5.40 pts
Code MigrationVals AI54.61%9of 379of 71Claude Sonnet 5.515.22 pts
CyberBench v1.1Vals AI70.42%15of 3118of 43GPT-6 Sol7.56 pts
EMBVals AI76.67%1of 371of 69Claude Fable 5.1—
Finance Agent (v2)Vals AI58.88%6of 377of 73Gemini 4 Argon6.52 pts
IOIVals AI90.78%6of 316of 38Gemini 4 Argon9.22 pts
Legal Research BenchVals AI55.29%1of 371of 72Muse Spark 1.3 Max—
MedCodeVals AI53.51%6of 358of 104Claude Opus 510.06 pts
MedScribeVals AI91.29%2of 362of 106Claude Opus 5.50.14 pts
Public Benefits Bench v1.1Vals AI74.90%2of 292of 45Claude Opus 52.03 pts
SAGEVals AI48.53%15of 3227of 90Claude Opus 4.77.57 pts
SRE BenchVals AI22.90%7of 157of 16GPT-6 Astra33.97 pts
Tax Agent BenchVals AI77.64%1of 371of 64Claude Fable 5.1—
GDPval-AA v2.1Artificial Analysis1,734.673of 42no list readClaude Opus 5.5111.5
AA-Briefcase v1.1Artificial Analysis1,678.373of 41no list readClaude Opus 5.5143.48
Humanity's Last ExamArtificial Analysis59.13%2of 42no list readClaude Opus 5.52.22 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%LiveCodeBench%GPQA Diamond%MMLU Pro%MMMU Pro%Vibe Code Bench v1.1%Terminal-Bench Science 0.1%ProgramBench%
Claude Sonnet 5.5Anthropic67.04%5663.64%––––92.39%53.33%–
Claude Opus 5.5Anthropic66.97%5859.60%––––90.29%59.05%18.50%
Claude Fable 5.1Anthropic65.83%5352.02%90.52%93.43%92.38%90.64%90.26%43.33%7.00%
Claude Opus 5Anthropic63.67%5148.99%89.03%93.43%91.59%89.88%88.40%28.57%3.00%
GPT-6 AstraOpenAI63.13%5359.09%––––89.59%63.33%5.50%
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
Claude Fable 5.1 (Adaptive Reasoning, Max Effort, Default Fallback)$7.6369.3 tok/s

Results by benchmark

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

BenchmarkIndependent runAs reported by a publisher
PhysicianBenchHealthcareno published result
61.0%PhysicianBenchAnthropic · 7 October 2026Rank 3 of 6 in this publisher table
Global MMLUMultilingual knowledgeno published result
94.0%Global MMLUAnthropic · 7 October 2026Rank 2 of 4 in this publisher table
MILUMultilingual knowledgeno published result
93.0%MILUAnthropic · 7 October 2026Rank 2 of 4 in this publisher table
τ³-BankingTool use47.22%Artificial Analysisnot published
ITBench-AAIT operations49.53%Artificial Analysisnot published
AA-AnalystAgentData analysis57.50%Artificial Analysisnot published
MLCR-AAMedical reasoning71.11%Artificial Analysisnot published
SWE-Bench Pro V2 FullRepository coding99.10%Scale AI / SEALnot published
SWE-Bench Pro V2 HARDRepository coding92.20%Scale AI / SEALnot published
Vals RSI IndexAutonomous research36.09%Vals AInot published
Vibe Code Bench 1–100Application development28.00%Vals AInot published
CUA-benchComputer use13.17%Vals AInot published
Harvey LAB-AALegal reasoning93.02%Artificial Analysisnot published
LABBench 2Scientific reasoningno published result
68.6%LABBench 2Google DeepMind · October 2026Rank 4 of 4 in this publisher table
RiemannBenchMathematicsno published result
65.6%RiemannBenchGoogle DeepMind · October 2026Rank 4 of 4 in this publisher table
GraphWalks (up to 128k)Long contextno published result
91.4%GraphWalks (Up to 128k, BFS F1)Google DeepMind · October 2026Rank 3 of 4 in this publisher table
GraphWalks (256k to 1M)Long contextno published result
65.0%GraphWalks (256k to 1M, BFS F1)Google DeepMind · October 2026Rank 4 of 4 in this publisher table
LVBenchVideo understandingno published result
79.7%LVBench (no tools)Google DeepMind · October 2026Rank 4 of 4 in this publisher table
CWE-bench v1Securityno published result
58.0%CWE-bench v1Google DeepMind · October 2026Rank 4 of 4 in this publisher table
AutomationBench-AAWorkflow automation59.38%Artificial Analysisnot published
SciCodeScientific coding63.08%Artificial Analysisnot published
GDP.pdfDocument reasoning26.20%Artificial Analysisnot published
CritPtScientific reasoning29.71%Artificial Analysisnot published
AA-OmniscienceFactual knowledge43.45Artificial Analysisnot published
AA-LCR v1.1Long-context reasoning85.33%Artificial Analysisnot published
Terminal-Bench 4.0Agentic coding52.02%Artificial Analysis
58.08%Vals AI
57.9%Terminal-Bench 4.0Google DeepMind · October 2026Rank 3 of 4 in this publisher table
55.8%Terminal-Bench 4.0OpenAI · 3 September 2026Rank 2 of 6 in this publisher table
55.8%Terminal-Bench 4.0Anthropic · 22 September 2026Rank 3 of 5 in this publisher table
Terminal-Bench 2.1Agentic coding85.02%Vals AI
91.39%Artificial Analysis
not published
LiveCodeBenchCode generation90.52%Vals AInot published
GPQA DiamondScience reasoning93.43%Vals AI
93.74%Artificial Analysis
93.7%GPQA DiamondOpenAI · 3 September 2026Rank 4 of 6 in this publisher table
MMLU ProKnowledge92.38%Vals AInot published
MMMU ProMultimodal reasoning90.64%Vals AInot published
Vibe Code Bench v1.1Product coding90.26%Vals AI
90.3%Vibe Code BenchGoogle DeepMind · October 2026Rank 2 of 4 in this publisher table
Terminal-Bench Science 0.1Agentic science43.33%Artificial Analysis
40.00%Vals AI
52.6%Terminal-Bench Science 0.1Google DeepMind · October 2026Rank 4 of 4 in this publisher table
52.6%Terminal-Bench Science 0.1OpenAI · 3 September 2026Rank 2 of 5 in this publisher table
52.6%Terminal-Bench-Science 0.1Anthropic · 22 September 2026Rank 3 of 5 in this publisher table
ProgramBenchAgentic coding7.00%Vals AInot published
ProofBench v1.1Mathematics100.00%Vals AInot published
LegalBenchProfessional knowledge88.51%Vals AInot published
Harvey's Legal Agent BenchmarkProfessional knowledge6.67%Vals AI
6.7%Harvey's Legal Agent BenchmarkGoogle DeepMind · October 2026Rank 2 of 4 in this publisher table
SkillsBenchAgentic tooling61.55%Vals AInot published
MysteryMechanismScientific reasoning47.75%Vals AInot published
Vals IndexComposite65.83%Vals AI
65.8%Vals IndexGoogle DeepMind · October 2026Rank 3 of 4 in this publisher table
Code MigrationAgentic coding54.61%Vals AInot published
CyberBench v1.1Security70.42%Vals AInot published
EMBFinancial modelling76.67%Vals AInot published
Finance Agent (v2)Financial analysis58.88%Vals AI
58.9%Vals Finance Agent v2Google DeepMind · October 2026Rank 2 of 4 in this publisher table
IOICompetitive programming90.78%Vals AInot published
Legal Research BenchProfessional knowledge55.29%Vals AInot published
MedCodeHealthcare administration53.51%Vals AInot published
MedScribeHealthcare administration91.29%Vals AInot published
Public Benefits Bench v1.1Public services74.90%Vals AInot published
SAGEEducation48.53%Vals AInot published
SRE BenchSecurity22.90%Vals AInot published
Tax Agent BenchProfessional knowledge77.64%Vals AInot published
OSWorld 2.0Computer useno published result
80.7%OSWorld 2.0 (partial)Anthropic · 22 September 2026Rank 2 of 3 in this publisher table
AutomationBenchBusiness workflowsno published result
31.4%AutomationBenchGoogle DeepMind · October 2026Rank 4 of 4 in this publisher table
31.4%AutomationBench 1.0.6OpenAI · 3 September 2026Rank 2 of 5 in this publisher table
31.4%AutomationBenchAnthropic · 22 September 2026Rank 3 of 5 in this publisher table
31.4%AutomationBench 1.0.6OpenAI · 29 September 2026Rank 5 of 5 in this publisher table
CursorBench 4.0Agentic codingno published result
51.8%CursorBench 4.0Anthropic · 22 September 2026Rank 2 of 4 in this publisher table
DeepSWE v1.1Software engineeringno published result
67.4%DeepSWE v1.1Google DeepMind · October 2026Rank 4 of 4 in this publisher table
67.4%DeepSWE v1.1OpenAI · 3 September 2026Rank 6 of 6 in this publisher table
FrontierCode v1.1Agentic codingno published result
50.9FrontierCode 1.1 MainOpenAI · 3 September 2026Rank 4 of 6 in this publisher table
50.3%FrontierCode v1.1 (Main)Anthropic · 22 September 2026Rank 3 of 5 in this publisher table
GDPval-AA v2.1Knowledge work1,734.67Artificial Analysis
1,735GDPval-AA v2.1Anthropic · 22 September 2026Rank 2 of 5 in this publisher table
AA-Briefcase v1.1Knowledge work1,678.37Artificial Analysisnot published
ChartographyVisual chart recognitionno published result
46.2%Chartography (no tools)Google DeepMind · October 2026Rank 4 of 4 in this publisher table
88.4%ChartographyAnthropic · 22 September 2026Rank 2 of 3 in this publisher table
Humanity's Last ExamMultidisciplinary reasoning59.13%Artificial Analysis
65.0%Humanity's Last Exam (with tools)OpenAI · 3 September 2026Rank 1 of 4 in this publisher table
65.6%Humanity's Last Exam (with tools)Anthropic · 22 September 2026Rank 2 of 4 in this publisher table
ARC-AGI-2Abstract reasoningno published result
90.0%ARC-AGI-2OpenAI · 3 September 2026Rank 4 of 5 in this publisher table
FrontierMath Tier 4 (v2)Mathematicsno published result
87.8%FrontierMath Tier 4 (v2)OpenAI · 3 September 2026Rank 3 of 5 in this publisher table
Artificial Analysis Intelligence IndexComposite53Artificial Analysisnot published
FrontierSWEAgentic researchno published result
56.3%FrontierSWE v2Google DeepMind · October 2026Rank 3 of 4 in this publisher table
PostTrainBenchAI researchno published result
40.2%PostTrainBench v1.1Google DeepMind · October 2026Rank 4 of 4 in this publisher table
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.