Skip to content

AnthropicFrontier

Claude Opus 5.5

First Claude 5.5 model: about Fable 5.1's level at 40% lower cost than Opus 5. Top of LMArena's Code Arena and Agent Arena.

Announced
Access
Public

Benchmark scores

TestScoreBest on the trackerHow it was measured
Epoch Capabilities Indexmany tests combined into one score169 pointsGPT-6.2 Astra 174 pointsIndependentlisted as "Claude Opus 5.5", CI 165.78 to 171.91Run 9 Nov 2026
AA Intelligence Index v5about ten tests in one score59.1 pointsClaude Fable 5.2 64.6 pointsIndependentv5.0.1; listed as Claude Opus 5.5 (Max, Default Fallback)Run 10 Nov 2026
Remote Labor Indexpaid freelance projects21%Claude Fable 5.2 41%IndependentOpus 5.5Run 7 Oct 2026
FrontierSWE V2very long engineering jobs62%Claude Fable 5.2 70%Independentproximus harness, max effort, mean@5 over 34 tasks, 20-hour budget (best@5 71.5)
Terminal-Bench 4.1long command-line tasks66%Top scoreIndependentClaude Code harness, max effort, 330 trials, ±3.11 (95% CI)
MirrorCoderebuilding whole programs79%Top scoreIndependenthigh reasoning effort, MirrorCode (ML, +Private, 2L): 15 programs x 2 languages, 3 runs each, 10B-token / 7-day budget, ±7.9 SERun 22 Sept 2026
APEX-Agentsprofessional office tasks74%Gemini 4 Carbon 84%Independentpass@1, max effort; listed as "Opus 5.5 [claude-opus-5-5]"
FrontierMath T1–3hard maths problems91%Claude Fable 5.2 95%Independentmax reasoning effort (claude-opus-5-5_max)Run 22 Sept 2026
GPQA DiamondPhD-level science questionsPhD experts: 70%Saturated91%Gemini 4 Argon 96%Independentmax reasoning effort (claude-opus-5-5_max)Run 22 Sept 2026
FrontierMath T4research-level maths problemsSaturated95%GPT-6.1 Sol 100%Independentmax reasoning effort (claude-opus-5-5_max)Run 22 Sept 2026
ARC-AGI-2visual puzzles, model aloneAverage person: 60%Saturated93%GPT-6 Astra 95%IndependentARC Prize config "Claude Opus 5.5 (High)" (anthropic-claude-opus-5-5-high)
OTIS Mock AIMEcompetition mathsSaturated100%Top scoreIndependentmax reasoning effort (claude-opus-5-5_max)Run 22 Sept 2026
METR time horizon (TH1.1 + TH2)task length an AI agent finishes half the time1.6 daysGPT-6 Astra 1.7 daysIndependentPooled TH1.1 + TH2 suites; 95% CI 714.9–75176.6 min

Compare every model on the same tests →

Sources