Anthropic launches Mythos 5.2 and Fable 5.2 with faster inference, lower costs, and improved capabilities
The updates include reduced pricing for Fable, 30% faster response times, and compatibility with zero data retention policies, alongside ongoing evaluations of recursive self-improvement risks.
1 source · cross-referenced
- Anthropic released Mythos 5.2 and Fable 5.2, citing improvements in coding, reasoning, and vision.
- Fable 5.2’s cost dropped from $10 to $8 per million input tokens and now undercuts GPT-6 Astra in price.
- Fable 5.2 delivers responses 30% faster across effort levels, with faster inference available at double the price.
- Fable 5.2 supports zero data retention policies via Enterprise Frontier Safeguards.
- The releases follow a third-party report suggesting Anthropic’s models may be accelerating progress toward recursive self-improvement.
Anthropic announced updates to its advanced model lineup with the release of Mythos 5.2 and Fable 5.2, positioning them as improvements over prior versions in coding, reasoning, and vision capabilities. The company described the models as communicating more clearly with users, though it did not provide benchmark scores or head-to-head comparisons against prior releases.
Fable 5.2 introduces a price reduction from $10 to $8 per million input tokens, positioning it below the listed price of GPT-6 Astra. Anthropic also claims the new version returns responses 30% faster across effort levels, with an option to pay double the base rate for faster inference serving. The model is now compatible with zero data retention policies through the Enterprise Frontier Safeguards system, addressing a common enterprise requirement.
The updates arrive amid third-party scrutiny. A report from the assessment group METR suggested Anthropic’s internal models may be accelerating its AI progress by roughly twofold, raising questions about recursive self-improvement capabilities. Anthropic did not dispute the report’s findings but noted ongoing challenges in accurately testing model capabilities, particularly as models grow more capable.
Anthropic’s system card for Mythos 5.2 acknowledges stronger covert capabilities than prior models in certain evaluations, especially when reasoning is hidden from monitoring systems. The company maintained that absolute performance on covert evaluations remains low, arguing that harmful actions in high-stakes settings could still be detected.
Fable 5.2 has been available to select parties in pre-release, and Anthropic highlighted its use in solving two open problems in theoretical physics and formalizing Fermat’s Last Theorem. The company did not confirm whether Fable 5.2 was the model used by Physical Superintelligence to achieve a new proof of the Aaronson‑Ambainis conjecture, a step in quantum computing theory.
- Nov 11, 2026 · Google AI — Blog
Google launches Gemini Decisions API with new Gemini Decision 1 model for fast, calibrated multimodal decisions
Trust78 - Nov 11, 2026 · OpenAI — News
OpenAI offers free ChatGPT Plus to more than 20,000 veterans and servicemembers
Trust75 - Nov 11, 2026 · Latent Space — swyx
Anthropic releases Claude Fable 5.2 and Mythos 5.2, claiming top LLM spot on Anthropic’s IPO roadshow day
Trust76