Claude Opus 5.5 Should Raise Your Ambitions

Anthropic introduced Claude Opus 5.5, the first model in what it calls the new Claude 5.5 family. According to the official pitch, it performs at the level of Claude Fable 5.1 on most work and costs 40% less to run than Opus 5. Anthropic highlights agentic coding, security and improved communications; early tester blurbs, flagged as AI generated, praise agentic coding skills, efficiency, readability and communication, long autonomous runs, reliability and describe it as a major upgrade.

Pricing is listed at $4/$20, 20% lower than Opus 5, with cache pricing of $0.20, 60% lower. Estimated overall costs typically drop 40% while speed rises 30%, and subscription limits were increased. A fast mode is available at $8/$40 with up to 2.5x the speed. Opus 5.5 is available with zero data retention. The report questions that availability given the model is described as at least as capable as Fable 5.1 and, per Anthropic, extremely strong on cyber tasks; guardrails are said to be similar to Fable 5.1.

On benchmarks, Opus 5.5 is described as almost universally better than every model except Astra, and usually above Astra. Artificial Analysis places it in the clear lead overall in intelligence at 58, and its five-point lead dominates Astra across most tracked benchmarks, with Astra's biggest lead in GDP.pdf. On the new Terminal-Bench-Science 0.1, Astra scores 63% and Opus 5.5 on xhigh 62%. WeirdML v3 has Astra at 42.2%, Opus 5.5 at 31.2% and Fable 5.1 at 26%.

Sam Bowman said Anthropic thinks Opus 5.5 is sufficiently safer than its predecessors that releasing it more likely than not reduces misalignment-related risks. OpenAI cut GPT-6 Sol prices 50% to $2/$10 and Luna to $0.10/$0.50; CEO Sam Altman called Sol and Luna big improvements over their 5.6-family predecessors. In testing at Box, Aaron Levie reported 63% fewer tokens used, 42% less verbosity and 30% faster versus Opus 5.