Anthropic’s New Claude Opus 5.5 Matches Fable at 60% the Price

Anthropic unveiled Claude Opus 5.5 as the newest member of its Claude 5.5 family, positioning it as a lower-cost alternative to its top-tier Fable 5.1 while claiming comparable performance on most tasks. The company says Opus 5.5 undercuts the model it replaces—Opus 5—by about 20% at default pricing and runs at roughly 60% of the per-token cost of Fable 5.1. Anthropic also lowered cache-read costs, which are important for long agentic sessions, to $0.20 per million tokens. On benchmarks published by Anthropic, Opus 5.5 posted improvements over Fable 5.1 and the prior Opus 5 in several agentic coding and knowledge-work evaluations. For Terminal-Bench 4.0, which measures the share of professional tasks an agent can complete in a command-line environment, Opus 5.5 scored 66.4% versus Fable 5.1’s 55.8% and GPT-6 Astra’s 57.9%. In FrontierCode, a pass-rate metric that simulates whether code changes would be accepted in a real pipeline, Opus 5.5 achieved 54.4% compared with Fable 5.1’s 50.3%. On GDPval-AA v2.1, an Elo-style ranking across 44 occupations, Opus 5.5 scored 1846 against Fable 5.1’s 1735 and Opus 5’s 1708. Opus 5.5 did not top every test Anthropic published. On AutomationBench, a Zapier-built benchmark evaluating end-to-end business workflows, GPT-6 Astra led with 41.4% versus Opus 5.5’s 40.0%. In scientific agentic work tested by Terminal-Bench-Science 0.1, Astra again outscored Opus 5.5, posting 64.6% against Opus 5.5’s 58.7%. Anthropic said Opus 5.5 carries the same cybersecurity and biology safeguards as its Mythos-class models, the restricted tier used for vetted cybersecurity and biology researchers, and that most cybersecurity tasks are still automatically routed to the older Opus 4.8 in its systems. The company has made Opus 5.5 available across its cloud partners including AWS, Google Cloud, and Microsoft Azure, and plans to roll out Sonnet 5.5 and Haiku 5.5 with similar price reductions in the coming weeks.