Anthropic has released two new flagship models: Claude Fable 5.1, available broadly through the Claude API, and Claude Mythos 5.1, which remains restricted to a select group of vetted users inside the company’s Project Glasswing program. Both models target the same class of demanding work that has become the real battleground in enterprise AI: long-horizon coding tasks, multistep research, and dense document processing.
The release landed on September 1, 2026, and builds directly on the Fable 5 and Mythos 5 lineage that Anthropic first shipped on June 9, 2026.
What the models actually do differently
Fable 5.1 and Mythos 5.1 each support one million tokens in context, with output capacity reaching 128,000 tokens per response.
Fable 5.1 adds two notable beta features that weren’t available in its predecessor. The first is mid-conversation effort adjustment, letting users dial up or down how hard the model works on a given step without starting a new session. The second is content provenance tracking, which tags outputs so downstream systems can verify where generated content originated.
Mythos 5.1 takes a different approach. For the approved participants inside Project Glasswing, Anthropic has removed the safety classifiers and fallback restrictions that govern Fable 5.1. Fable 5.1 retains safety classifiers alongside fallback options for high-risk tasks, keeping it suitable for broad commercial deployment.
Pricing, context, and the June suspension
The pricing structure for Fable 5.1 holds steady against its predecessor at $10 per million input tokens and $50 per million output tokens. Cache read costs have dropped to 25% of prior levels.
In June 2026, Anthropic temporarily suspended access to the earlier Fable 5 and Mythos 5 models following U.S. export control actions tied to cybersecurity concerns. Access was restored by the end of that month.
What this means for the competitive landscape
Market analysts indicate the timing of these releases aligns with increasing competition in the AI sector, particularly with the impending launch of OpenAI’s GPT-6.
The cache read cost reduction in Fable 5.1 is a concrete financial consideration. At scale, the difference between full-price and 25%-of-prior-price cache reads on a one-million-token context window adds up across millions of API calls per month.
The content provenance tracking feature also deserves attention beyond its compliance optics. As AI-generated content becomes harder to distinguish from human-produced work, having a verifiable chain of origin embedded in outputs could become table stakes for regulated industries including finance, healthcare, and legal services. Anthropic is treating that as a native capability rather than an add-on.