Skip to content

Anthropic

Claude Opus 5

Close to Fable 5 at half the price. Top of Scale AI's SWE-Bench Pro V2 (99.4%) and of OSWorld 2.0 (44.3%, scored pass or fail).

Announced
Access
Public

Benchmark scores

TestScoreBest on the trackerHow it was measured
Epoch Capabilities Indexmany tests combined into one score164 pointsGPT-6.2 Astra 174 pointsIndependentlisted as "Claude Opus 5", CI 162.02 to 167.3Run 9 Nov 2026
AA Intelligence Index v5about ten tests in one score44.2 pointsClaude Fable 5.2 64.6 pointsIndependentv5.0.1; listed as Claude Opus 5 (Max)Run 10 Nov 2026
FrontierSWE V2very long engineering jobs52%Claude Fable 5.2 70%Independentproximus harness, max effort, mean@5 over 34 tasks, 20-hour budget (best@5 60.7)
Terminal-Bench 4.1long command-line tasks53%Claude Opus 5.5 66%IndependentClaude Code harness, xhigh effort, 330 trials, ±3.09 (95% CI)
APEX-Agentsprofessional office tasks66%Gemini 4 Carbon 84%Independentpass@1, max effort; listed as "Opus 5"
ARC-AGI-3puzzle games, standard harnessTypical first-time human player: 100%30%Claude Fable 5.2 81%IndependentARC Prize config "Claude Opus 5 (High)" (anthropic-claude-opus-5-high), standard harness
FrontierMath T1–3hard maths problems86%Claude Fable 5.2 95%Independentmax reasoning effort (claude-opus-5_max)Run 24 Jul 2026
OSWorld-Verifiedusing a computerPeople (original task set): 72%83%Claude Fable 5 86%Independentplain model, max 100 steps, 1M-context variant ("claude-opus-5[1m]", 300.2 / 360 tasks)Run 1 Aug 2026
GPQA DiamondPhD-level science questionsPhD experts: 70%Saturated94%Gemini 4 Argon 96%Independentmax reasoning effort (claude-opus-5_max)Run 24 Jul 2026
FrontierMath T4research-level maths problemsSaturated73%GPT-6.1 Sol 100%Independentmax reasoning effort (claude-opus-5_max)Run 24 Jul 2026
ARC-AGI-2visual puzzles, model aloneAverage person: 60%Saturated90%GPT-6 Astra 95%IndependentARC Prize config "Claude Opus 5 (Max)" (anthropic-claude-opus-5-max)
OTIS Mock AIMEcompetition mathsSaturated99%GPT-5.5 100%Independentmax reasoning effort (claude-opus-5_max)Run 24 Jul 2026
METR time horizon (TH1.1 + TH2)task length an AI agent finishes half the time20 hGPT-6 Astra 1.7 daysIndependentPooled TH1.1 + TH2 suites; 95% CI 566.8–5174.8 min

Compare every model on the same tests →

Also on these charts

Sources