Claude Opus 5.5

Caylent's analysis of Anthropic's Claude Opus 5.5 launch highlights gains in coding and professional knowledge work alongside a 20% price cut to $4 per million input tokens and $20 per million output tokens. The model can reportedly match or exceed Fable 5.1's results on selected evaluations while charging 60% less per Representative results from Anthropic's launch evaluations include Terminal-Bench 4.0 at 66.4% (vs. 52.3% for Opus 5, 55.8% for Fable 5.1), FrontierCode v1.1 Main at 54.4% (vs. 48.0% and 50.3%), GDPval-AA v2.1 at 1,846 Elo (vs. 1,708 and 1,735), and OSWorld 2.0 partial at 81.8% (vs. 74.0% and 80.7%). Practical API chan

CodeRabbit's AI code-review pipeline evaluation of Anthropic's Claude Opus 5.5 finds the model edging past the production baseline on bug coverage in an open-source test and posting larger gains in both coverage and precision on a harder subset, with the trade-off being more comments and higher reported token usage. Th Specific changes versus Opus 5 include stronger coding performance at lower effort (Opus 5.5 at medium effort matched or beat Opus 5 at high effort on multi-step coding tasks using roughly half the tokens), a 20% list-price reduction (input from $5 to $4, output from $25 to $20 per million tokens), and a 60% drop in ca

Sonar's independent evaluation of Claude Opus 5.5 (run at high effort against the same Java benchmark used for every model in its leaderboard, covering HumanEval, MBPP, and ComplexCodeEval) records an 87.7% pass rate on the 544 HumanEval and MBPP tasks with executable tests, within one percentage point of Opus 5's 88.6 Total findings fell 42% from 18,814 to 10,941, with code smell density down 21% and vulnerability density down 9%. Blocker-level findings dropped across categories: reliability blockers down 41%, security blockers down 53%, and maintainability blockers down 20%. The tradeoffs noted are a 12% rise in bug density (576 to

Anthropic launched Claude Opus 5.5 on September 22, 2026, positioning it as comparable to its top-tier Fable 5.1 while costing 40% less to run than its predecessor. Reuters reports the model underwent external testing by AI safety groups Frontier Design and METR before launch and includes safeguards previously reserved Pricing is set at $4 per million input tokens and $20 per million output tokens, 20% below Opus 5. According to Anthropic, Opus 5.5 outscored OpenAI's GPT-5.6 Sol on a software development benchmark while costing roughly one-third as much to run, and is available on Amazon Web Services, Google Cloud, and Microsoft Azur