Claude Opus 5.5 Released: Matches Fable 5.1 on Most Tasks, Typical Cost Drops 40%

动察 Beating AI News Flash: Anthropic releases Claude Opus 5.5, the first model in the Claude 5.5 series. The company says it reaches Fable 5.1 levels on most tasks, but costs 40% less than Opus 5 on typical workloads and delivers output more than 30% faster. API pricing drops to $4 input and $20 output per million tokens, both 20% lower than Opus 5; cache reads fall to $0.20, 60% cheaper.

Coding benchmarks show the most notable gains. On Terminal-Bench 4.0, Opus 5.5 scores 66.4%, above GPT-6 Astra's 57.9%; on FrontierCode v1.1 it is 54.4% versus 53.3%. Anthropic says that when running FrontierCode at default effort, Opus 5.5 exceeds Astra's highest score at about one-fifth the per-question cost; on GDPval-AA v2.1 it also leads with 1846 Elo, above Fable 5.1's 1735 and Astra's 1542.

But it does not lead across the board. On AutomationBench, GPT-6 Astra scores 41.4% versus Opus 5.5's 40.0%; on Terminal-Bench-Science, Astra also scores higher.

This is the first model released after Anthropic announced it would "slow down frontier model progress." The company says Opus 5.5 achieved its best result yet in its automated behavior audit. The five-hour quotas for Pro, Max, and Team are also being raised in tandem, and Sonnet 5.5 and Haiku 5.5 will launch in the coming weeks.