Anthropic releases Claude Opus 5.5 at lower cost than Opus 5

Anthropic released Claude Opus 5.5 on September 22, 2026, pricing the model at $4 per million input tokens and $20 per million output tokens through its API. Anthropic said the price is about 40% lower than Opus 5 on typical workloads.

Cached input reads cost $0.20 per million tokens, which Anthropic said is 60% less than under Opus 5, and cache writes cost $5 per million tokens. Anthropic also offers a “fast mode” tier priced at $8 per million input tokens and $40 per million output tokens.

The model is available immediately on the Claude Platform, Amazon Web Services, Google Cloud and Microsoft Azure, Anthropic said.

Anthropic said Opus 5.5 generates output more than 30% faster than Opus 5. The company said the new model matches or exceeds the performance of Claude Fable 5.1 — which Anthropic released on September 2 — on most benchmarks it tested. Anthropic cited a score of 66.4% on Terminal-Bench 4.0 for Opus 5.5 compared with 55.8% for Fable 5.1, and 1846 Elo on GDPval-AA v2.1 compared with 1735 Elo for Fable 5.1. On biology- and cybersecurity-related evaluations, Anthropic said Opus 5.5 is comparable to Claude Mythos 5.1, a separate, more restricted Anthropic model, and is being deployed with safeguards similar to those on Fable 5.1. The benchmark figures are Anthropic’s own; the company did not point to independent evaluation of them alongside the release.

On safety, Anthropic said Opus 5.5 is the strongest-performing model it has tested on its automated behavioral audit. The company said the model attempted to circumvent its own operating boundaries about 85% less often than Opus 5 or Claude Mythos 5.1, and that it matches or beats Opus 5 on prompt injection resistance in every setting tested. On a separate benchmark run by the security firm Gray Swan, Anthropic said Opus 5.5 tied Fable 5.1 for the lowest prompt injection success rate of any model tested. Anthropic did not publish the behavioral-audit methodology alongside the release.