Anthropic rolls out Claude Opus 4.8 and teases broader Mythos release in coming weeks

Anthropic launched Claude Opus 4.8 on Thursday, a new version of its flagship AI model with improved coding, reasoning, and agentic workflow capabilities, while also announcing that its more advanced Mythos class models will be released in the coming weeks. The model is available at the same standard pricing as its predecessor and includes new features such as dynamic workflows in Claude Code and user-controlled effort settings. Anthropic said Opus 4.8 is roughly four times less likely than Opus 4.7 to leave unacknowledged flaws in code and showed substantially lower rates of misaligned behavior in internal assessments.

Anthropic rolls out Claude Opus 4.8 and teases broader Mythos release in coming weeks Claude Opus 4.8 adds stronger coding and workflow tools as Anthropic says Mythos class models are coming in weeks. Anthropic launched https://www.anthropic.com/news/claude-opus-4-8 Claude Opus 4.8 on Thursday, rolling out a new version of its flagship AI model with stronger performance across coding, agentic tasks, reasoning, and practical knowledge work, while also teasing a broader release of its more advanced Mythos class models in the coming weeks. The model builds on Claude Opus 4.7 and is available today at the same standard price, with regular usage priced at $5 per million input tokens and $25 per million output tokens. Developers can access the model through the Claude API using claude-opus-4-8. Anthropic said Opus 4.8 is designed to be a more effective collaborator, with early testers reporting stronger judgment, better tool use, and improved reliability on long running agentic workflows. The company said one of the model’s clearest improvements is honesty. Opus 4.8 is more likely to flag uncertainty in its own work and less likely to claim progress without enough evidence. Anthropic said its evaluations showed Opus 4.8 was around four times less likely than Opus 4.7 to leave flaws in code it had written without noting them. Anthropic also said its alignment assessment found that Opus 4.8 showed substantially lower rates of misaligned behavior, including deception or cooperation with misuse, compared with Opus 4.7. The company said the model performed similarly to Claude Mythos Preview, its best aligned model. The launch comes with several product updates. Claude Code is getting dynamic workflows, a research preview feature that lets Claude plan large tasks and run hundreds of parallel subagents in a single session before verifying its own outputs. Anthropic said the feature can support codebase scale migrations across hundreds of thousands of lines of code, using existing test suites as a benchmark. Claude users are also getting new effort controls on claude.ai http://claude.ai and Cowork, allowing them to decide how much effort Claude should put into a response. Higher settings make the model think more deeply, while lower settings produce faster answers and use rate limits more slowly. Opus 4.8 defaults to high effort, which Anthropic said offers the best balance between quality and user experience. Users can also choose extra effort, called xhigh in Claude Code, or max effort for harder tasks and long running workflows. The company also reduced pricing for fast mode, where Opus 4.8 can work at 2.5 times the speed. Fast mode is now three times cheaper than it was for previous models, priced at $10 per million input tokens and $50 per million output tokens. Anthropic also updated the Messages API to allow system entries inside the messages array. The change lets developers adjust Claude’s instructions mid task without breaking prompt caching or routing the update through a user turn, which can help agents update permissions, token budgets, or environment context as they run. Anthropic described Opus 4.8 as a modest but tangible upgrade over Opus 4.7. The company said it is also working on lower cost models with similar capabilities and plans to bring Mythos class models to customers in the coming weeks after developing stronger cybersecurity safeguards. Disclosure: This article was edited by Estefano Gomez. For more information on how we create and review content, see our Editorial Policy https://cryptobriefing.com/editorial-policy/ .