Anthropic releases Opus 5.5 with lower prices and Fable-level performance
Anthropic has released Opus 5.5, claiming it outperforms Fable in many tests. Output tokens now cost $20 per million, while one test migrated 680,000 lines of code in under a day.
Image: itc.ua · Author: \u0428\u0430\u0434\u0440\u0456\u043d \u0410\u043d\u0434\u0440\u0456\u0439 · source articleEditorial excerpt for news reporting
Anthropic has released Opus 5.5, the highest tier in its three-level Claude lineup. The company says the new model outperforms the larger Fable model in many tests and completed informal tasks that Fable could not.
Output tokens cost $20 per million, down from $25 for the previous model. Anthropic also reports faster operation and lower computing requirements.
The update uses less jargon and more often places important information at the start of its responses. The launch came two months after Opus 5 was released on July 24.
Programming and performance tests
An early tester used Opus 5.5 to migrate a 680,000-line codebase in under a day. Anthropic estimates that an engineering team could have needed several weeks for the same work.
The model also reduced a web application’s page-load time successfully in 39 of 40 attempts. In a test migrating HAProxy from C to Rust, Opus 5.5 finished in 9.5 hours, compared with 12 hours for Fable 5.1, while costing 51% less.
Claude Code mode and safeguards
A fast mode for Opus 5.5 in Claude Code and Claude Platform can accelerate work by up to 2.5 times. It has separate pricing: $8 per million input tokens and $40 per million output tokens.
Anthropic says the model matched Opus 5’s lowest prompt-injection success rate across all tested scenarios, including programming, tool use, computer interaction and web browsing. In a Gray Swan test, the result matched Fable 5.1.
Because Anthropic says Opus 5.5 has capabilities in biology and cybersecurity, it is covered by the same safeguards as Fable. These restrictions include limiting exploit searches in compiled programs and the development of detectable biological weapons.
Safety training was broadly similar to that used for earlier models and included alignment testing and pre-release evaluations by METR and Frontier Design. Anthropic said improved safety and monitoring systems are being prepared for future models. Sonnet 5.5 and Haiku 5.5 are planned for release “in the coming weeks.”
The release is Anthropic’s first model launch since CEO Dario Amodei supported slowing the development of advanced systems so safety methods can keep pace with their capabilities.
What we know
- Anthropic says Opus 5.5 outperformed Fable in many tests.
- Output-token pricing fell to $20 per million from $25 for the previous model.
- In the HAProxy test, Opus 5.5 took 9.5 hours, 2.5 hours less than Fable 5.1.
- The model migrated a 680,000-line codebase in under a day.
- Sonnet 5.5 and Haiku 5.5 are due in the coming weeks.
Trust: Single source
- The story relies on one publisher. A second independent confirmation has not been established from the cited sources.
- Sources: ITC.ua.
COMMUNITY
Discussion
Sign in to join the discussion.
Loading comments…