Anthropic has launched Claude Opus 5.5, the first model in its new 5.5 series. The company claims the model matches the performance of its flagship Claude Fable 5.1 in most tasks while offering significant cost reductions for users.
The new model is 20% cheaper to run than its predecessor, Opus 5, and 60% less expensive than Fable 5.1. Input tokens are priced at $4 per million, while output tokens cost $20 per million. Cache reads have seen a 60% price drop to $0.20 per million tokens.
In benchmark testing, Opus 5.5 achieved a 66.4% success rate on Terminal-Bench 4.0, surpassing Fable 5.1 at 55.8%. On the GDPval-AA v2.1 ranking, the model scored 1846 points, compared to 1735 for Fable 5.1. However, it trails behind GPT-6 Astra on AutomationBench and Terminal-Bench-Science 0.1.
This release follows a recent essay by CEO Dario Amodei, who advocated for a slower pace in AI capability development to prioritize safety. Despite this, Opus 5.5 incorporates the same safety protocols as Fable 5.1, though specific cybersecurity tasks remain routed to the older Opus 4.8 model.
Claude Opus 5.5 is currently available across Amazon Web Services, Google Cloud, and Microsoft Azure. Anthropic plans to release the Sonnet 5.5 and Haiku 5.5 models in the coming weeks, extending the new pricing structure to the rest of its product lineup.
