Anthropic launched its latest AI model, Claude Opus 4.8, on Thursday and announced the upcoming release of its more powerful Claude Mythos model [1, 2].
The update signals a push toward higher reliability and autonomy in AI assistants, aiming to reduce hallucinations while increasing the efficiency of complex coding and reasoning tasks.
Anthropic said the Opus 4.8 model is designed to be a more effective collaborator with improved multidisciplinary reasoning [3, 4]. A spokesperson for the company said, "Claude Opus 4.8 is our most honest model yet" [3]. The company's product team said the model possesses "sharper judgement, more honesty about its progress, and the ability to work independently for longer than its predecessors" [5].
Regarding pricing, Opus 4.8 is available at $5 per million input tokens and $25 per million output tokens [6]. These rates are the same as those for the previous Opus 4.7 model [6]. However, the new release introduces a fast-mode option that is three times cheaper than the previous fast mode [7].
While Opus 4.8 is available now, the company is preparing for a larger shift. Anthropic said it is working to release the "market-moving" Mythos model to all customers in the coming weeks [1]. This upcoming model is expected to perform at a higher level than the current Opus series [2].
The rollout of Opus 4.8 on May 28, 2026, serves as a bridge to the Mythos class [1, 8]. The company said these iterations will provide a more honest and capable tool for developers, and enterprise users [3, 4].
“"Claude Opus 4.8 is our most honest model yet."”
Anthropic is employing a tiered release strategy by deploying an incremental update (Opus 4.8) to stabilize performance and cost-efficiency before launching a major architectural leap with the Mythos model. By emphasizing 'honesty' and 'independent work,' the company is attempting to solve the persistent AI problem of hallucinations, positioning itself as the reliable alternative for high-stakes professional and technical environments.
