Skip to content

Anthropic

Claude Opus 4.8

Opus upgrade at the same price. Launched with a setting in claude.ai for how much effort the model spends.

Announced
Access
Public

Benchmark scores

TestScoreBest on the trackerHow it was measured
Epoch Capabilities Indexmany tests combined into one score160 pointsGPT-6.2 Astra 174 pointsIndependentlisted as "Claude Opus 4.8", CI 157.82 to 162.25Run 9 Nov 2026
Remote Labor Indexpaid freelance projects8.3%Claude Fable 5.2 41%IndependentOpus 4.8Run 1 Jul 2026
Terminal-Bench 4.1long command-line tasks24%Claude Opus 5.5 66%IndependentClaude Code harness, max effort, 330 trials, ±3.59 (95% CI)
APEX-Agentsprofessional office tasks49%Gemini 4 Carbon 84%Independentpass@1, max effort; listed as "Opus 4.8"
ARC-AGI-3puzzle games, standard harnessTypical first-time human player: 100%1.5%Claude Fable 5.2 81%IndependentARC Prize config "Claude Opus 4.8 (High)" (anthropic-opus-4-8-high), standard harness
FrontierMath T1–3hard maths problems80%Claude Fable 5.2 95%Independentmax reasoning effort (claude-opus-4-8_max)Run 10 Jun 2026
GPQA DiamondPhD-level science questionsPhD experts: 70%Saturated91%Gemini 4 Argon 96%Independentmax reasoning effort (claude-opus-4-8_max)Run 7 Jun 2026
FrontierMath T4research-level maths problemsSaturated56%GPT-6.1 Sol 100%Independentmax reasoning effort (claude-opus-4-8_max)Run 10 Jun 2026
ARC-AGI-2visual puzzles, model aloneAverage person: 60%Saturated72%GPT-6 Astra 95%IndependentARC Prize config "Claude Opus 4.8 (High)" (anthropic-opus-4-8-high)
OTIS Mock AIMEcompetition mathsSaturated98%GPT-5.5 100%Independentmax reasoning effort (claude-opus-4-8_max)Run 7 Jun 2026
METR time horizon (TH1.1 + TH2)task length an AI agent finishes half the time12 hGPT-6 Astra 1.7 daysIndependentPooled TH1.1 + TH2 suites; 95% CI 366.9–1268.1 min

Compare every model on the same tests →

Also on these charts

Sources