Anthropic has released Claude Opus 5.5, which the company claims outperforms Fable 5.1 and is 40 percent cheaper. Above all, performance, security, and communication style are said to have improved compared to Opus 5 and Fable 5.1. In addition, the smaller and more affordable Sonnet 5.5 and Haiku 5.5 models are set to be released sometime in the coming weeks.
Once again, an Opus model outperforms Fable in many benchmarks. However, Opus 5 proved to be more limited than Fable in real-world use, particularly when dealing with complex issues. The added value of 5.5 is said to revolve around its cost. Anthropic never shares details about the sizes of its own LLMs, but we can assume that Opus 5.5 is significantly more compact (and therefore cheaper to run) than the Fable models.
Migrating entire codebases in a day
Opus 5.5 is said to be a strong coder. In fact, according to Anthropic, an entire codebase can be migrated in a single day. An early tester managed to complete this in three hours with a codebase of 200,000 lines. For a rewrite task, Opus 5.5 finished in 9.5 hours, while Fable 5.1 took 12 hours; the former was 51 percent cheaper.
Opus 5.5 is the first model Anthropic has released since it, along with other AI labs, pledged to roll out the most powerful LLMs more slowly. Development of Opus 5.5 was already well underway before these joint plans were announced. Nevertheless, the emphasis was apparently already on security. The new model is said to perform as well as or better than Opus 5 when it comes to fending off prompt injection attacks.
Fewer “Claudisms”
Anyone who regularly interacts with AI chatbots knows that they have a unique, and not always desirable, writing style. Opus 5.5 is said to have taken feedback on Opus 5 to heart in this area; the most important information is at the top, and the model uses jargon or peculiar language less frequently.
Read also: The AI race isn’t slowing down