AIEvals

Models

Claude Sonnet 5

The preceding Claude Sonnet release, retained for comparison with Sonnet 5.5.

AnthropicReleased 30 June 2026Context 1M tokensOutput 128k tokensPrivate weights$2.00 in / $10.00 out per Mtok

Family Claude Sonnet 560 benchmarks with a published resultNo first place on the independent columns 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 AI51.77%20of 3321of 41Gemini 4 Argon17.13 pts
Artificial Analysis Intelligence IndexArtificial Analysis3833of 42no list readClaude Opus 5.520 pts
Terminal-Bench 4.0Artificial Analysis14.14%28of 40no list readClaude Sonnet 5.549.50 pts
SWE-bench (Verified)Vals AI79.60%19of 2325of 88Claude Opus 517.40 pts
LiveCodeBenchVals AI82.43%18of 2153of 143Claude Fable 5.18.09 pts
GPQA DiamondVals AI88.89%19of 2332of 138Gemini 3.1 Pro6.56 pts
MMLU ProVals AI87.55%16of 2426of 138Claude Fable 5.14.83 pts
MMMU ProVals AI83.01%17of 2035of 93Claude Fable 5.17.63 pts
Vibe Code Bench v1.1Vals AI81.33%18of 3819of 106Claude Sonnet 5.511.06 pts
ProgramBenchVals AI0.00%19of 2923of 54Claude Opus 5.518.50 pts
τ³-BankingArtificial Analysis37.32%21of 30no list readQwen3.8 Max14.02 pts
AA-AnalystAgentArtificial Analysis46.25%8of 18no list readGemini 3.7 Flash13.75 pts
MLCR-AAArtificial Analysis55.00%6of 37no list readClaude Sonnet 5.520.00 pts
SWE-Bench Pro V2 FullScale AI / SEAL93.15%8of 10no list readClaude Opus 56.25 pts
SWE-Bench Pro V2 HARDScale AI / SEAL88.20%4of 10no list readClaude Opus 59.80 pts
Vibe Code Bench 1–100Vals AI13.82%16of 2016of 20Claude Opus 5.516.54 pts
Harvey LAB-AAArtificial Analysis90.07%12of 20no list readMuse Spark 1.3 Max5.41 pts
EnterpriseOps-Gym-AAArtificial Analysis44.67%11of 22no list readClaude Fable 56.45 pts
AutomationBench-AAArtificial Analysis36.51%35of 40no list readGemini 4 Argon41.00 pts
SciCodeArtificial Analysis54.28%26of 40no list readClaude Opus 5.512.62 pts
GDP.pdfArtificial Analysis13.20%33of 40no list readGPT-6.1 Sol17.80 pts
CritPtArtificial Analysis16.86%32of 42no list readGPT-5.6 Sol15.43 pts
AA-OmniscienceArtificial Analysis16.4523of 42no list readClaude Opus 5.529.97
AA-LCR v1.1Artificial Analysis82.00%19of 42no list readKimi K36.67 pts
Terminal-Bench 2.1Vals AI74.53%18of 3518of 75Claude Opus 5.513.11 pts
ProofBench v1.1Vals AI77.00%12of 3414of 46Claude Sonnet 5.523.00 pts
LegalBenchVals AI83.92%21of 2744of 149Claude Fable 54.64 pts
Harvey's Legal Agent BenchmarkVals AI5.00%25of 3731of 73Muse Spark 1.220.42 pts
SkillsBenchVals AI46.48%17of 2026of 35DeepSeek V4.1 Flash23.32 pts
Vals Multimodal IndexVals AI68.83%8of 178of 33Claude Fable 55.32 pts
Code MigrationVals AI44.39%20of 3720of 71Claude Sonnet 5.525.44 pts
CyberBench v1.1Vals AI61.91%24of 3131of 43GPT-6 Sol16.07 pts
EMBVals AI66.32%19of 3719of 69Claude Fable 5.110.35 pts
Finance Agent (v2)Vals AI53.91%20of 3725of 73Gemini 4 Argon11.49 pts
IOIVals AI45.00%26of 3127of 38Gemini 4 Argon55.00 pts
Legal Research BenchVals AI41.83%19of 3720of 72Muse Spark 1.3 Max13.46 pts
MedCodeVals AI47.54%18of 3531of 104Claude Opus 516.03 pts
MedScribeVals AI76.05%35of 3675of 106Claude Opus 5.515.38 pts
Public Benefits Bench v1.1Vals AI66.03%17of 2917of 45Claude Opus 510.90 pts
SAGEVals AI48.92%13of 3224of 90Claude Opus 4.77.18 pts
Tax Agent BenchVals AI62.27%24of 3728of 64Claude Fable 5.115.37 pts
DeepSWE v1.1Datacurve53.85%18of 2018of 28GPT-6 Astra20.27 pts
GDPval-AA v2.1Artificial Analysis1,449.228of 42no list readClaude Opus 5.5396.97
AA-Briefcase v1.1Artificial Analysis1,359.1926of 41no list readClaude Opus 5.5462.66
Humanity's Last ExamArtificial Analysis41.29%32of 42no list readClaude Opus 5.520.06 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%SWE-bench%LiveCodeBench%GPQA Diamond%MMLU Pro%MMMU Pro%Vibe Code Bench v1.1%ProgramBench%
GPT-5.6 TerraOpenAI53.09%4235.35%95.40%85.93%90.91%86.66%86.47%74.59%0.50%
Grok 4.6xAI52.09%4417.17%95.60%88.22%94.70%89.40%–76.24%–
Claude Sonnet 5Anthropic51.77%3814.14%79.60%82.43%88.89%87.55%83.01%81.33%0.00%
GPT-5.6 LunaOpenAI51.69%3711.62%93.00%–91.67%86.04%85.03%77.06%0.00%
DeepSeek V4.1 FlashDeepSeek51.32%3926.77%–––––84.74%–
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 Sonnet 5 (Adaptive Reasoning, Max Effort)$5.0978.7 tok/s

Results by benchmark

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

BenchmarkIndependent runAs reported by a publisher
BenchCAD Vision2Code (no tools)CAD reasoningno published result
0.322BenchCAD Vision2Code (no tools)Anthropic · 7 October 2026Rank 4 of 5 in this publisher table
BenchCAD Vision2Code (with tools)CAD reasoningno published result
0.519BenchCAD Vision2Code (with tools)Anthropic · 7 October 2026Rank 4 of 5 in this publisher table
SpatialBench VerifiedLife sciencesno published result
68.9%SpatialBench VerifiedAnthropic · 7 October 2026Rank 1 of 2 in this publisher table
SingleCellBenchLife sciencesno published result
56.5%SingleCellBenchAnthropic · 7 October 2026Rank 1 of 2 in this publisher table
Medicinal chemistry (ADME)Life sciencesno published result
41.3%Medicinal chemistry (ADME)Anthropic · 7 October 2026Rank 4 of 4 in this publisher table
Protein design (sequence generation)Life sciencesno published result
20.2%Protein design (sequence generation)Anthropic · 7 October 2026Rank 3 of 3 in this publisher table
Protein design (library ranking)Life sciencesno published result
44.0%Protein design (library ranking)Anthropic · 7 October 2026Rank 3 of 3 in this publisher table
Biomedical image analysisLife sciencesno published result
42.3%Biomedical image analysisAnthropic · 7 October 2026Rank 2 of 2 in this publisher table
Protocols (troubleshooting)Life sciencesno published result
49.9%Protocols (troubleshooting)Anthropic · 7 October 2026Rank 3 of 3 in this publisher table
Protocols (understanding V2)Life sciencesno published result
58.9%Protocols (understanding V2)Anthropic · 7 October 2026Rank 3 of 3 in this publisher table
τ³-BankingTool use37.32%Artificial Analysisnot published
AA-AnalystAgentData analysis46.25%Artificial Analysisnot published
MLCR-AAMedical reasoning55.00%Artificial Analysisnot published
SWE-Bench Pro V2 FullRepository coding93.15%Scale AI / SEALnot published
SWE-Bench Pro V2 HARDRepository coding88.20%Scale AI / SEALnot published
Vibe Code Bench 1–100Application development13.82%Vals AInot published
Harvey LAB-AALegal reasoning90.07%Artificial Analysisnot published
EnterpriseOps-Gym-AAEnterprise workflows44.67%Artificial Analysisnot published
AutomationBench-AAWorkflow automation36.51%Artificial Analysisnot published
SciCodeScientific coding54.28%Artificial Analysisnot published
GDP.pdfDocument reasoning13.20%Artificial Analysisnot published
CritPtScientific reasoning16.86%Artificial Analysisnot published
AA-OmniscienceFactual knowledge16.45Artificial Analysisnot published
AA-LCR v1.1Long-context reasoning82.00%Artificial Analysisnot published
Terminal-Bench 4.0Agentic coding14.14%Artificial Analysis
9.60%Vals AI
10.3%Terminal-Bench 4.0Anthropic · 28 September 2026Rank 3 of 3 in this publisher table
Terminal-Bench 2.1Agentic coding74.53%Vals AI
80.52%Artificial Analysis
not published
SWE-bench (Verified)Software engineering79.60%Vals AInot published
LiveCodeBenchCode generation82.43%Vals AInot published
GPQA DiamondScience reasoning88.89%Vals AI
91.11%Artificial Analysis
not published
MMLU ProKnowledge87.55%Vals AInot published
MMMU ProMultimodal reasoning83.01%Vals AI
77.28%Artificial Analysis
not published
Vibe Code Bench v1.1Product coding81.33%Vals AInot published
Terminal-Bench Science 0.1Agentic science
2.86%Vals AI
not published
ProgramBenchAgentic coding0.00%Vals AInot published
ProofBench v1.1Mathematics77.00%Vals AInot published
LegalBenchProfessional knowledge83.92%Vals AInot published
Harvey's Legal Agent BenchmarkProfessional knowledge5.00%Vals AInot published
SkillsBenchAgentic tooling46.48%Vals AInot published
Vals Multimodal IndexComposite68.83%Vals AInot published
Vals IndexComposite51.77%Vals AInot published
Code MigrationAgentic coding44.39%Vals AInot published
CyberBench v1.1Security61.91%Vals AInot published
EMBFinancial modelling66.32%Vals AInot published
Finance Agent (v2)Financial analysis53.91%Vals AInot published
IOICompetitive programming45.00%Vals AInot published
Legal Research BenchProfessional knowledge41.83%Vals AInot published
MedCodeHealthcare administration47.54%Vals AInot published
MedScribeHealthcare administration76.05%Vals AInot published
Public Benefits Bench v1.1Public services66.03%Vals AInot published
SAGEEducation48.92%Vals AInot published
Tax Agent BenchProfessional knowledge62.27%Vals AInot published
OSWorld 2.1Computer useno published result
57.0%OSWorld 2.1 (partial)Anthropic · 28 September 2026Rank 3 of 3 in this publisher table
CursorBench 4.0Agentic codingno published result
34.1%CursorBench 4.0Anthropic · 28 September 2026Rank 3 of 3 in this publisher table
DeepSWE v1.1Software engineering53.85%Datacurvenot published
FrontierCode v1.1Agentic codingno published result
42.4%FrontierCode 1.1 (Main)Anthropic · 28 September 2026Rank 4 of 4 in this publisher table
GDPval-AA v2.1Knowledge work1,449.2Artificial Analysis
1,449GDPval-AA v2.1Anthropic · 28 September 2026Rank 4 of 4 in this publisher table
AA-Briefcase v1.1Knowledge work1,359.19Artificial Analysis
1,359AA-Briefcase v1.1Anthropic · 28 September 2026Rank 4 of 4 in this publisher table
ChartographyVisual chart recognitionno published result
15.6%Chartography (no tools)Anthropic · 7 October 2026Rank 9 of 10 in this publisher table
73.2%Chartography (with tools)Anthropic · 7 October 2026Rank 4 of 5 in this publisher table
15.6%Chartography (no tools)Anthropic · 28 September 2026Rank 4 of 4 in this publisher table
Humanity's Last ExamMultidisciplinary reasoning41.29%Artificial Analysis
54.9%Humanity's Last Exam (with tools)Anthropic · 28 September 2026Rank 3 of 3 in this publisher table
Artificial Analysis Intelligence IndexComposite38Artificial 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.