Anthropic Launches Claude Opus 4.8

Anthropic released Claude Opus 4.8 on May 28, a new flagship AI model that the company says outperforms competitors from OpenAI, Google, and Meta across major reasoning and coding benchmarks.

The model arrives roughly five months after Claude 4.5 Sonnet and marks a return to the Opus line, which Anthropic had previously paused to work on efficiency and safety. Chief product officer Mike Krieger said Opus 4.8 was trained with a focus on extended reasoning tasks, allowing it to handle multi-step problems that earlier versions struggled with. In a blog post, Anthropic claimed the model scored 94.3 percent on the MATH-500 benchmark and surpassed GPT-5 on several graduate-level science tests.

Opus 4.8 also introduces a new feature called "extended thinking mode," which lets the model spend more time on difficult queries before responding. The feature is optional and can be toggled on by users who need deeper analysis. It is priced at $20 per million input tokens and $80 per million output tokens, making it Anthropic's most expensive model to date.

Early access partners including Notion and DuckDuckGo reported improved performance on complex document summarization and multi-turn conversations. The model is available starting today through Anthropic's API and major cloud partners.