Anthropic to Release Mythos-Class AI Model to All Customers Within Weeks
The general release of Mythos, a model class above Opus and already sought by Japanese financial institutions, signals that Anthropic believes its safety measures are now sufficient for wider deployment of frontier AI.
Reporting from 1 sources: ASCII.jp.
Anthropic announced Claude Opus 4.8 on May 28 and said its higher-intelligence Mythos-class model, currently in preview for cybersecurity use, will become generally available within a few weeks. Opus 4.8 improves coding, reasoning, and honesty, with reduced defect rates. Japanese megabanks are expected to start using Mythos this month.
Anthropic released Claude Opus 4.8 on May 28 and announced that its Mythos-class model, currently in a preview phase for cybersecurity organizations, will become available to all customers within the next few weeks. The company described Mythos as a new class with higher intelligence than Opus, and said safeguards to detect and block dangerous outputs had been developed before expanding access. Opus 4.8 shows improvements in coding, agentic tasks, reasoning, and knowledge work benchmarks, with pricing unchanged at $5 per million input tokens and $25 per million output tokens. Anthropic highlighted a reduction in the model's tendency to produce false confident answers: the probability of leaving unnoticed defects in written code is about one-quarter that of the previous generation. The AI coding tool Claude Code also introduced dynamic workflows as a research preview, allowing parallel sub-agents to run in a single session. Japanese megabanks Mitsubishi UFJ, Sumitomo Mitsui, and Mizuho are expected to begin using Mythos within May.
Synthesized by Yomimono from the 1 cited source below, including Japanese-language reporting where cited, then editorially reviewed before publishing.