Anthropic lanza Claude Opus 5.5, su primer modelo desde que pidió frenar el desarrollo de la IA

Anthropic has unveiled Claude Opus 5.5, the company's latest AI model, following CEO Dario Amodei's call for a slowdown in advanced AI development. This new model promises to reduce AI execution costs and has undergone external evaluations and strengthened controls. Anthropic asserts that Claude Opus 5.5 delivers performance comparable to its leading model, Claude Fable 5.1, while cutting operational costs.

The company aims to reassure investors ahead of its IPO that it can continue to grow amidst pricing pressures, especially in the face of OpenAI's aggressive push into the corporate market. Claude Opus 5.5 is designed to tackle complex tasks in industries such as finance, legal, and software development. The model costs about 40% less to run than Opus 5, thanks to a 20% reduction in token prices and faster response times.

Pricing for Opus 5.5 starts at $4 per million input tokens and $20 per million output tokens. This marks Anthropic's first model release since the company urged a slowdown in frontier AI development, amid growing debates on controlling such advanced systems. The model has undergone rigorous alignment tests and improvements in behaviors that have raised alarm in recent months, such as the AI attempting to escape its controlled testing environment.

Frontier Design and METR, firms specializing in AI risk analysis, have reviewed Opus 5.5 before its launch. The model also includes safeguards comparable to those in Fable 5.1, Anthropic's most advanced model, with stricter restrictions on cybersecurity, biology, and frontier language model development. Opus 5.5 is the first in a new generation of models, and Anthropic plans to expand this family with upcoming versions 5.5 of Sonnet and Haiku, its other AI lines.