Anthropic launches Claude Opus 4.8, prepares wider access to Mythos model

Anthropic pushed Claude Opus 4.8 into wider use at the same price as its prior Opus generation, while keeping its more powerful Mythos model under tight control, a split strategy that underscores how fiercely frontier-AI labs are competing on capability, cost and trust.

The new Opus model has a 1 million-token context window and is available to Claude Pro, Max, Team and Enterprise users, as well as through the Claude Platform and cloud partners Amazon Web Services, Google Cloud and Microsoft Foundry. Anthropic said pricing starts at $5 per million input tokens and $25 per million output tokens, with prompt caching cutting costs by up to 90 percent and batch processing by 50 percent. For U.S.-only workloads, the company said U.S.-only inference is available at 1.1 times pricing.

Anthropic said Opus 4.8 performs better on coding, agentic tasks, reasoning and professional work, and gives users more control over how much effort the model spends on a task. It also added new features to Claude Code, including dynamic workflows designed to handle larger problems and a fast mode that runs at 2.5 times the speed and, in Anthropic’s words, three times cheaper than previous models. Early testers said the model felt more reliable, sharper in judgment and better at signaling uncertainty instead of confidently inventing progress, an improvement that matters in a market where large language models are often criticized for sounding authoritative when they are wrong.

Mythos remains the more sensitive release. Anthropic described Mythos Preview as a controlled defensive-security initiative, not a general public rollout, and said launch partners include Amazon Web Services, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, the Linux Foundation, Microsoft, NVIDIA and Palo Alto Networks. The company said it extended access to more than 40 additional organizations that build or maintain critical software infrastructure, and said that roughly 50 partners had used Mythos Preview to find more than 10,000 high- or critical-severity vulnerabilities. Anthropic later revised its policy so those cybersecurity findings can be shared with others facing similar vulnerabilities, but it still does not plan a general release.

The timing adds commercial pressure to the product news. Anthropic said on May 28, 2026, that it raised $65 billion in a Series H round at a $965 billion post-money valuation, backed by Altimeter Capital, Dragoneer, Greenoaks and Sequoia Capital. That valuation move put Anthropic ahead of OpenAI on a valuation basis, while leaving the company with a familiar challenge: prove that each new model is not only more capable, but also safer, cheaper and more useful in the real world.