Anthropic Launches Claude Opus 5.5 With Fable-Level Performance at 40% Lower Cost

Anthropic has shipped Claude Opus 5.5, the first model in its new Claude 5.5 family and a release the company says performs at the level of Claude Fable 5.1 on most work while costing 40% less to run than the previous Opus 5. It is also Anthropic's first frontier release since it called for pacing the frontier, and the first Opus-class model to ship with Fable-style safety classifiers.

The headline numbers

Opus 5.5 lands two months after Opus 5 (released July 24) with an unusual combination: more capability at a lower price. Input and output tokens are priced at $4 and $20 per million respectively, 20% below Opus 5, while cache reads drop 60% to $0.20 per million tokens. Anthropic says the model generates output more than 30% faster, and that using fewer tokens per task nets out to a roughly 40% drop in real-world cost.

The model ships with a 1M-token context window and 128K maximum output (up to 300K on the Batch API in beta), with adaptive thinking always on at a default "medium" effort. Its knowledge cutoff is June 2026.

What it is good at

Anthropic positions Opus 5.5 as a long-horizon agentic coding and knowledge-work model. Compared with Opus 5, the system card reports gains in coding, agentic and computer-use tasks, mathematical and scientific reasoning, and long-horizon professional work. On many evaluations it matches or exceeds both Claude Fable 5.1 and Claude Mythos 5.1.

The standout is biology. Anthropic says Opus 5.5 is highly capable in biology, exceeding Opus 5 and matching or beating Mythos 5.1 across many areas of work, while concluding it still sits at the CB-1 capability level (synthesis of non-novel weapons) rather than crossing the CB-2 threshold. The model also scores as the strongest Anthropic has tested on its automated behavioral audit, the company's most comprehensive alignment test.

A safer Opus

For the first time, an Opus-class model ships with safety classifiers similar to Claude Fable 5.1 across biology, cybersecurity, and AI development, so requests are refused more frequently than in previous Opus versions. The model was tested before release by external evaluators including Frontier Design and METR.

Migration watch-outs

Four breaking changes affect code running on Opus 5. Thinking can no longer be disabled, forced tool use returns an error, thinking blocks are now tied to the model and the conversation, and the earlier computer_20251124 computer-use tool is not accepted on the Claude API or Google Cloud. Builders who stream text between tool calls should also note that inter-tool text now returns inside thinking blocks, empty at the default display setting, so progress updates can go quiet unless a display value is set.

Why it matters for builders

Opus 5.5 matters less for raw benchmark leadership and more for economics. A Fable-class reasoning model at roughly 40% lower cost changes the unit economics of long-running agentic and coding workloads, where cache reads dominate the bill:

Model	Context	Max output	Price (in / out per MTok)	Thinking
Claude Fable 5.1	1M	128K	$10 / $50	Adaptive (high)
Claude Opus 5.5	1M	128K	$4 / $20	Adaptive (medium)
Claude Sonnet 5	1M	128K	$2 / $10	Adaptive (high)

Anthropic says Sonnet 5.5 and Haiku 5.5 follow "in the coming weeks," which would extend the same efficiency gains down the model ladder.

Key takeaway: Opus 5.5 is Anthropic's efficiency play: Fable-class reasoning at a 40% lower cost, with Fable-grade safety classifiers. For builders running long-horizon agentic or coding workloads, the 60% cheaper cache reads are the number that matters most.