AIEvals

Models

Claude Opus 5.5

Anthropic’s flagship Claude 5.5 model for demanding coding, research and knowledge work.

AnthropicReleased 22 September 2026Context 1M tokensOutput 128k tokensPrivate weights$4.00 in / $20.00 out per Mtok

Family Claude 5.565 benchmarks with a published result12 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 AI66.97%3of 333of 41Gemini 4 Argon1.93 pts
Artificial Analysis Intelligence IndexArtificial Analysis581of 42no list readClaude Opus 5.5—
Terminal-Bench 4.0Artificial Analysis59.60%2of 40no list readClaude Sonnet 5.54.04 pts
Vibe Code Bench v1.1Vals AI90.29%4of 384of 106Claude Sonnet 5.52.10 pts
Terminal-Bench Science 0.1Artificial Analysis59.05%2of 24no list readGPT-6 Astra4.28 pts
ProgramBenchVals AI18.50%1of 291of 54Claude Opus 5.5—
ITBench-AAArtificial Analysis38.23%16of 18no list readGPT-5.6 Sol17.98 pts
MLCR-AAArtificial Analysis66.67%3of 37no list readClaude Sonnet 5.58.33 pts
Vals RSI IndexVals AI37.31%1of 201of 23Claude Opus 5.5—
Vibe Code Bench 1–100Vals AI30.36%1of 201of 20Claude Opus 5.5—
CUA-benchVals AI14.00%2of 82of 8GPT-6 Astra5.17 pts
Harvey LAB-AAArtificial Analysis91.20%9of 20no list readMuse Spark 1.3 Max4.28 pts
AutomationBench-AAArtificial Analysis69.54%3of 40no list readGemini 4 Argon7.97 pts
SciCodeArtificial Analysis66.90%1of 40no list readClaude Opus 5.5—
GDP.pdfArtificial Analysis26.20%5of 40no list readGPT-6.1 Sol4.80 pts
CritPtArtificial Analysis31.71%2of 42no list readGPT-5.6 Sol0.58 pts
AA-OmniscienceArtificial Analysis46.421of 42no list readClaude Opus 5.5—
AA-LCR v1.1Artificial Analysis84.67%5of 42no list readKimi K34.00 pts
Terminal-Bench 2.1Vals AI87.64%1of 351of 75Claude Opus 5.5—
ProofBench v1.1Vals AI100.00%1of 341of 46Claude Sonnet 5.5—
Harvey's Legal Agent BenchmarkVals AI3.75%27of 3734of 73Muse Spark 1.221.67 pts
MysteryMechanismVals AI49.55%2of 212of 21GPT-6 Astra3.60 pts
BioMysteryBenchVals AI79.26%3of 193of 21Claude Sonnet 5.51.85 pts
Code MigrationVals AI66.65%4of 374of 71Claude Sonnet 5.53.18 pts
CyberBench v1.1Vals AI55.36%26of 3135of 43GPT-6 Sol22.62 pts
EMBVals AI75.94%2of 372of 69Claude Fable 5.10.73 pts
Finance Agent (v2)Vals AI58.59%8of 379of 73Gemini 4 Argon6.81 pts
IOIVals AI95.06%4of 314of 38Gemini 4 Argon4.94 pts
Legal Research BenchVals AI50.48%5of 375of 72Muse Spark 1.3 Max4.81 pts
MedCodeVals AI49.80%10of 3517of 104Claude Opus 513.77 pts
MedScribeVals AI91.43%1of 361of 106Claude Opus 5.5—
Public Benefits Bench v1.1Vals AI70.64%3of 293of 45Claude Opus 56.29 pts
SAGEVals AI45.83%22of 3237of 90Claude Opus 4.710.27 pts
SRE BenchVals AI33.59%4of 154of 16GPT-6 Astra23.28 pts
Tax Agent BenchVals AI70.50%8of 379of 64Claude Fable 5.17.14 pts
GDPval-AA v2.1Artificial Analysis1,846.171of 42no list readClaude Opus 5.5—
AA-Briefcase v1.1Artificial Analysis1,821.851of 41no list readClaude Opus 5.5—
Humanity's Last ExamArtificial Analysis61.35%1of 42no list readClaude Opus 5.5—

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%
Gemini 4 ArgonGoogle68.90%5357.07%91.91%–2.50%
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%90.26%43.33%7.00%
Claude Opus 5Anthropic63.67%5148.99%88.40%28.57%3.00%
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
Claude Opus 5.5 (Adaptive Reasoning, Max Effort, Default Fallback)$5.9892.2 tok/s

Results by benchmark

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

BenchmarkIndependent runAs reported by a publisher
ProgramBench (hidden test pass rate)Long-context codingno published result
91.2%ProgramBench (hidden test pass rate)Anthropic · 7 October 2026Rank 1 of 3 in this publisher table
DRACO (Anthropic harness)Deep researchno published result
87.4%DRACO (Anthropic harness)Anthropic · 7 October 2026Rank 1 of 4 in this publisher table
WANDR (frozen index)Wide researchno published result
72.3%WANDR (frozen index)Anthropic · 7 October 2026Rank 1 of 4 in this publisher table
BenchCAD Vision2Code (no tools)CAD reasoningno published result
0.73BenchCAD Vision2Code (no tools)Anthropic · 7 October 2026Rank 2 of 5 in this publisher table
BenchCAD Vision2Code (with tools)CAD reasoningno published result
0.962BenchCAD Vision2Code (with tools)Anthropic · 7 October 2026Rank 2 of 5 in this publisher table
OSWorld 2.1 (offline strict pass)Computer useno published result
53.2%OSWorld 2.1 (offline strict pass)Anthropic · 7 October 2026Rank 1 of 6 in this publisher table
PhysicianBenchHealthcareno published result
68.4%PhysicianBenchAnthropic · 7 October 2026Rank 1 of 6 in this publisher table
Global MMLUMultilingual knowledgeno published result
94.3%Global MMLUAnthropic · 7 October 2026Rank 1 of 4 in this publisher table
MILUMultilingual knowledgeno published result
93.1%MILUAnthropic · 7 October 2026Rank 1 of 4 in this publisher table
Medicinal chemistry (ADME)Life sciencesno published result
63.5%Medicinal chemistry (ADME)Anthropic · 7 October 2026Rank 2 of 4 in this publisher table
ITBench-AAIT operations38.23%Artificial Analysisnot published
MLCR-AAMedical reasoning66.67%Artificial Analysisnot published
Vals RSI IndexAutonomous research37.31%Vals AInot published
Vibe Code Bench 1–100Application development30.36%Vals AInot published
CUA-benchComputer use14.00%Vals AInot published
Harvey LAB-AALegal reasoning91.20%Artificial Analysisnot published
LABBench 2Scientific reasoningno published result
73.1%LABBench 2Google DeepMind · October 2026Rank 3 of 4 in this publisher table
RiemannBenchMathematicsno published result
69.6%RiemannBenchGoogle DeepMind · October 2026Rank 3 of 4 in this publisher table
GraphWalks (up to 128k)Long contextno published result
90.6%GraphWalks (Up to 128k, BFS F1)Google DeepMind · October 2026Rank 4 of 4 in this publisher table
GraphWalks (256k to 1M)Long contextno published result
66.8%GraphWalks (256k to 1M, BFS F1)Google DeepMind · October 2026Rank 3 of 4 in this publisher table
LVBenchVideo understandingno published result
83.7%LVBench (no tools)Google DeepMind · October 2026Rank 3 of 4 in this publisher table
CWE-bench v1Securityno published result
67.0%CWE-bench v1Google DeepMind · October 2026Rank 3 of 4 in this publisher table
AutomationBench-AAWorkflow automation69.54%Artificial Analysisnot published
SciCodeScientific coding66.90%Artificial Analysisnot published
GDP.pdfDocument reasoning26.20%Artificial Analysisnot published
CritPtScientific reasoning31.71%Artificial Analysisnot published
AA-OmniscienceFactual knowledge46.42Artificial Analysisnot published
AA-LCR v1.1Long-context reasoning84.67%Artificial Analysisnot published
Terminal-Bench 4.0Agentic coding59.60%Artificial Analysis
65.15%Vals AI
66.4%Terminal-Bench 4.0Anthropic · 7 October 2026Rank 2 of 6 in this publisher table
66.4%Terminal-Bench 4.0Google DeepMind · October 2026Rank 1 of 4 in this publisher table
66.4%Terminal-Bench 4.0Anthropic · 22 September 2026Rank 1 of 5 in this publisher table
66.4%Terminal-Bench 4.0Anthropic · 28 September 2026Rank 2 of 3 in this publisher table
Terminal-Bench 2.1Agentic coding87.64%Vals AInot published
MMMU ProMultimodal reasoning
87.69%Artificial Analysis
not published
Vibe Code Bench v1.1Product coding90.29%Vals AI
90.3%Vibe Code BenchGoogle DeepMind · October 2026Rank 2 of 4 in this publisher table
Terminal-Bench Science 0.1Agentic science59.05%Artificial Analysis
47.14%Vals AI
58.7%Terminal-Bench-Science 0.1Anthropic · 7 October 2026Rank 2 of 5 in this publisher table
63.3%Terminal-Bench Science 0.1Google DeepMind · October 2026Rank 2 of 4 in this publisher table
58.7%Terminal-Bench-Science 0.1Anthropic · 22 September 2026Rank 2 of 5 in this publisher table
63.3%Terminal-Bench Science 0.1OpenAI · 29 September 2026Rank 2 of 4 in this publisher table
ProgramBenchAgentic coding18.50%Vals AInot published
ProofBench v1.1Mathematics100.00%Vals AInot published
Harvey's Legal Agent BenchmarkProfessional knowledge3.75%Vals AI
3.8%Harvey's Legal Agent BenchmarkGoogle DeepMind · October 2026Rank 4 of 4 in this publisher table
MysteryMechanismScientific reasoning49.55%Vals AInot published
Vals IndexComposite66.97%Vals AI
67.0%Vals IndexGoogle DeepMind · October 2026Rank 2 of 4 in this publisher table
BioMysteryBenchAgentic science79.26%Vals AInot published
Code MigrationAgentic coding66.65%Vals AInot published
CyberBench v1.1Security55.36%Vals AInot published
EMBFinancial modelling75.94%Vals AInot published
Finance Agent (v2)Financial analysis58.59%Vals AI
58.6%Vals Finance Agent v2Google DeepMind · October 2026Rank 3 of 4 in this publisher table
IOICompetitive programming95.06%Vals AInot published
Legal Research BenchProfessional knowledge50.48%Vals AInot published
MedCodeHealthcare administration49.80%Vals AInot published
MedScribeHealthcare administration91.43%Vals AInot published
Public Benefits Bench v1.1Public services70.64%Vals AInot published
SAGEEducation45.83%Vals AInot published
SRE BenchSecurity33.59%Vals AInot published
Tax Agent BenchProfessional knowledge70.50%Vals AInot published
OSWorld 2.0Computer useno published result
81.8%OSWorld 2.0 (partial)Anthropic · 22 September 2026Rank 1 of 3 in this publisher table
OSWorld 2.1Computer useno published result
87.2%OSWorld 2.1 (offline partial)Anthropic · 7 October 2026Rank 1 of 6 in this publisher table
81.8%OSWorld 2.1 (partial)Anthropic · 28 September 2026Rank 1 of 3 in this publisher table
AutomationBenchBusiness workflowsno published result
42.5%AutomationBenchGoogle DeepMind · October 2026Rank 2 of 4 in this publisher table
40.0%AutomationBenchAnthropic · 22 September 2026Rank 2 of 5 in this publisher table
42.5%AutomationBench 1.0.6OpenAI · 29 September 2026Rank 1 of 5 in this publisher table
CursorBench 4.0Agentic codingno published result
57.8%CursorBench 4.0Anthropic · 22 September 2026Rank 1 of 4 in this publisher table
57.8%CursorBench 4.0Anthropic · 28 September 2026Rank 1 of 3 in this publisher table
DeepSWE v1.1Software engineeringno published result
74.2%DeepSWE v1.1Google DeepMind · October 2026Rank 2 of 4 in this publisher table
FrontierCode v1.1Agentic codingno published result
54.4%FrontierCode v1.1 (Main)Anthropic · 22 September 2026Rank 1 of 5 in this publisher table
54.4%FrontierCode 1.1 (Main)Anthropic · 28 September 2026Rank 1 of 4 in this publisher table
GDPval-AA v2.1Knowledge work1,846.17Artificial Analysis
1,846GDPval-AA v2.1Anthropic · 22 September 2026Rank 1 of 5 in this publisher table
1,846GDPval-AA v2.1Anthropic · 28 September 2026Rank 1 of 4 in this publisher table
AA-Briefcase v1.1Knowledge work1,821.85Artificial Analysis
1,822AA-Briefcase v1.1Anthropic · 28 September 2026Rank 1 of 4 in this publisher table
ChartographyVisual chart recognitionno published result
64.4%Chartography (no tools)Anthropic · 7 October 2026Rank 2 of 10 in this publisher table
89.0%Chartography (with tools)Anthropic · 7 October 2026Rank 2 of 5 in this publisher table
66.3%Chartography (no tools)Google DeepMind · October 2026Rank 3 of 4 in this publisher table
89.0%ChartographyAnthropic · 22 September 2026Rank 1 of 3 in this publisher table
64.4%Chartography (no tools)Anthropic · 28 September 2026Rank 1 of 4 in this publisher table
Humanity's Last ExamMultidisciplinary reasoning61.35%Artificial Analysis
67.7%Humanity's Last Exam (with tools)Anthropic · 22 September 2026Rank 1 of 4 in this publisher table
67.7%Humanity's Last Exam (with tools)Anthropic · 28 September 2026Rank 1 of 3 in this publisher table
Agents' Last ExamComputer useno published result
38.2%Agent's Last Exam (binary pass rate)Google DeepMind · October 2026Rank 2 of 3 in this publisher table
Artificial Analysis Intelligence IndexComposite58Artificial Analysisnot published
FrontierSWEAgentic researchno published result
62.3%FrontierSWE v2Google DeepMind · October 2026Rank 2 of 4 in this publisher table
PostTrainBenchAI researchno published result
49.3%PostTrainBench v1.1Google DeepMind · October 2026Rank 1 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.