Anthropic launches Claude Opus 5.5 with Fable-level performance, 40% lower cost, now on GitHub and AWS

Anthropic released Claude Opus 5.5 on September 22, the first model in a new Claude 5.5 family, saying it performs at the level of its higher-tier Claude Fable 5.1 model on most work while costing about 40% less to run than Opus 5. The pricing detail underneath that headline is the part worth reading closely. API rates are $4 per million input tokens and $20 per million output tokens, against $5 and $25 for Opus 5, which is a 20% per-token cut, not 40%. Cache reads drop to $0.20 per million, 60% cheaper than Opus 5, and it is those cache savings that carry the 40% figure on the cache-heavy agentic workloads Anthropic describes as typical.

On the company's own benchmark suite, Opus 5.5 scored 1,846 on GDPval-AA v2.1, a test spanning real-world tasks across 44 occupations, against 1,735 for Fable 5.1 and 1,708 for Opus 5, and it generates output more than 30% faster than Opus 5. Anthropic attached two caveats that the launch coverage largely skipped: it says benchmark margins have become a less reliable guide at this level of capability, and that the real-world gap to Fable 5.1 is narrower than the scores suggest. The competitive picture is also split rather than one-sided. Opus 5.5 posted 66.4% on Terminal-Bench 4.0 against 57.9% for OpenAI's GPT-6 Astra, but Astra still leads Terminal-Bench-Science 0.1 at 64.6% to Opus 5.5's 58.7%.

Distribution arrived with the model rather than after it. Claude Opus 5.5 is available on Amazon Bedrock at AMZN, on Google Cloud at GOOGL and on Microsoft Azure at MSFT, and it shipped the same day into GitHub Copilot for Pro+, Max, Business and Enterprise subscribers. Anthropic said Claude Sonnet 5.5 and Claude Haiku 5.5 will follow in the coming weeks. For scale on the price gap it is competing into, GPT-6 Astra lists at $10 per million input tokens and $50 per million output tokens.

The launch lands ten days after Amodei published an essay urging AI companies to pace the frontier and slow development for safety reasons, which makes Opus 5.5 the first Anthropic model released under that stated posture. The company says outside evaluators including METR and Frontier Design tested the model before release, and that because Opus 5.5 is comparable to Mythos 5.1 in biology and cybersecurity it ships with safeguards modelled on Fable 5.1's: most cybersecurity tasks are rerouted to Opus 4.8, and broader biology access runs through a new Life Sciences Verification Program.

For anyone buying inference rather than selling it, the thing to watch is where the 40% actually shows up. A workload that reuses long contexts could capture most of it through the cheaper cache reads, while a low-cache workload sees closer to the 20% token cut, so the saving is a function of how an application is built rather than a flat discount. The Bloomberg item in this cluster frames the launch as pre-listing positioning for Anthropic, which would fit a price move aimed at holding enterprise volume. Whether cheaper, faster frontier capacity pulls more of that volume onto the three cloud platforms carrying it is the readable signal over the next few quarters.