Claude Opus 5.5

Claude Opus 5.5, the first model in Anthropic’s new Claude 5.5 family, has been launched as the company’s most capable and cost‑effective offering to date. The model matches the performance of Claude Fable 5.1 on most tasks while costing roughly 40 % less to run than its predecessor, Opus 5, on typical workloads. External evaluators such as Frontier Design and METR tested the system before release, and Anthropic’s own automated behavioral audit—its most comprehensive alignment test—recorded Opus 5.5 as the strongest‑performing model it has ever evaluated. Early adopters reported dramatic productivity gains: one tester completed a 680,000‑line code migration in under a day, a job that would have required weeks of engineering effort, and the model successfully optimized load times on 39 of 40 web‑app pages, outperforming Opus 5, which made only modest improvements that altered app behavior. Additional tests showed Opus 5.5 building a game from a single prompt with superior graphics and polish compared with all other Claude models.

Beyond raw performance, Opus 5.5 delivers notable safety and efficiency advances. The model achieved the highest scores to date on Anthropic’s alignment suite, showing a markedly lower propensity for hard‑to‑reverse actions, reduced likelihood of operating outside prescribed boundaries, and greater resistance to prompt‑injection attacks than recent releases. Safety testing now includes longer and impossible tasks as well as scenarios modeled on real incidents, though Anthropic acknowledges remaining limits. Cost reductions stem from lower compute requirements: input and output tokens are priced at $4 and $20 per million respectively—20 % cheaper than Opus 5—while cache reads, which dominate agentic and coding expenses, drop to $0.20 per million tokens, a 60 % cut. The model also generates output over 30 % faster, and Anthropic has raised five‑hour usage limits across Pro, Max, Team, and Enterprise plans, adding a flexible rate‑limit reset for subscribers.

The rollout of Opus 5.5 is accompanied by expanded access programs and forthcoming model variants. Verified organizations can now apply to the Life Sciences Verification Program to employ Opus 5.5 for biology research, and a Cyber Verification Program will open in the coming weeks for certified cybersecurity practitioners. Anthropic plans to release Claude Sonnet 5.5 and Claude Haiku 5.5 shortly, each incorporating similar performance, efficiency, and safety improvements. Benchmarks show Opus 5.5 leading in agentic coding, computer use, and knowledge work, though Anthropic notes that at this capability level benchmark margins become less reliable indicators of real‑world differences, with the practical gap between Opus 5.5 and Claude Fable 5.1 appearing narrower in internal use. When production safeguards intervene, the model may defer to earlier versions—Claude Opus 4.8 for cybersecurity tasks and Claude Opus 5 for biology and frontier LLM development—potentially tempering its headline performance figures but reinforcing safety priorities.

⚡ Effects Interpreter