ChainThink News, June 10: According to an official announcement, Anthropic has launched its new flagship large model, Claude Fable 5—the first Mythos-level model open to the public—with full API and enterprise edition availability now live.

The model is priced at $10 per million input tokens and $50 per million output tokens, representing a reduction of over 50% compared to the beta version. A parallel architecture-upgraded variant, Claude Mythos 5, is also launching, targeting security-focused institutions with fewer safety restrictions.

In performance benchmarks, Claude Fable 5 achieved a 29.3% score on the FrontierCode Diamond difficulty evaluation—significantly surpassing GPT-5.5’s 5.7% and Claude Opus’s 13.4%.

Stripe testing demonstrates that it can migrate a 50-million-line Ruby codebase in a single day—previously requiring a team two months to complete the same workload.

Beyond coding, the model has successfully passed the Pokémon FireRed game via screenshot input, reconstructed web source code, predicted adenovirus capsid assembly through reasoning alone—outperforming specialized protein models—and independently completed single-cell genomics research, designing a classification model that exceeds those published in Science, with only 1% of the parameters.

To mitigate misuse of frontier capabilities, Anthropic has deployed an independent safety classifier; high-risk queries automatically trigger fallback to Claude Opus 4.8, affecting less than 5% of sessions on average.

Enterprise traffic data will be retained for 30 days, with all manual reviews logged. Glasswing partners may lift the cybersecurity restrictions on Claude Mythos 5, while biomedical restrictions will be gradually opened to select institutions via trusted access mechanisms.

Subscription users may freely trial the model until June 22; thereafter, usage will require consumption points, pending future compute scalability for reintegration into standard subscriptions.