Claude Opus 5.5

Anthropic released Claude Opus 5.5 on September 22 2026 as the first model in a new Claude 5.5 family. The company says it runs about 40 % cheaper than the prior Opus 5 and generates output more than 30 % faster. Early testers reported a 680,000‑line code migration completed in less than a day, a task that would normally require weeks for a human team. The model also achieved the highest scores to date on Anthropic’s automated behavioral audit, a comprehensive alignment test. The performance jump positions Opus 5.5 as the leading model for agentic coding, computer use and knowledge work, according to Anthropic’s internal benchmarks. Lower compute demands translate into token pricing of $4 per million for input and $20 per million for output, both 20 % less than Opus 5, while cache reads cost $0.20 per million tokens, 60 % less. Safety improvements, such as reduced susceptibility to prompt injection and tighter guardrails for biology and cybersecurity, broaden the model’s applicability to regulated domains. Anthropic notes that benchmark margins are becoming less reliable at this capability level, and the real‑world gap between Opus 5.5 and Claude Fable 5.1 may be narrower than scores suggest. The company acknowledges limits in its alignment testing, especially for longer or impossible tasks, and external verification of those limits remains pending. Details of how safeguard interventions affect performance on specialized biology or frontier LLM tasks have not been fully disclosed.