Anthropic, the AI giant led by Dario Amodei, unveiled its latest AI model, Claude Fable 5.2, on Monday. The Fable series was released alongside Mythos 5.2, Anthropic’s state-of-the-art AI model designed specifically for cybersecurity. The new Mythos model is available to select cybersecurity professionals through a verification program. Both models reportedly offer improved performance in areas such as coding, reasoning, and visual capabilities. According to Anthropic, the new Fable and Mythos models also demonstrate superior communication skills compared to their predecessors. Furthermore, both models are more affordably priced than the previous Fable 5.1 and Mythos 5.1 versions.
Pricing and Availability of Claude Fable 5.2 and Mythos 5.2
In a blog post, Anthropic announced that the new Claude Fable 5.2 and Mythos 5.2 AI models are priced at $8 per million input tokens and $40 per million output tokens. In “fast” mode, the pricing is $16 per million input tokens and $80 per million output tokens.
The Claude Fable 5.2 AI model is currently available via Amazon Web Services, Google Cloud, Microsoft Azure, and other platforms; developers can also access it through the Claude platform. Additionally, Anthropic has increased usage limits for paid subscribers this week. Meanwhile, Mythos 5.2 is available to select cybersecurity professionals through a “Cyber Verification Program” and is accessible to enterprise users via Claude Security.
Finally, eligible users can access Claude Fable 5.2 and Claude Mythos 5.2—featuring a “fast mode”—via Claude Code, the Claude Platform, Claude models on Google Cloud, and Microsoft Foundry.
Capabilities of Claude Fable 5.2 and Mythos 5.2
The AI lab led by Dario Amodei claims that Claude Fable 5.2 and Mythos 5.2 offer performance improvements across several areas. Both AI models reportedly deliver output speeds up to 30% faster. Additionally, the “fast mode” can further accelerate token generation by up to 2.5 times. Fable 5.2 outperforms Google’s frontier AI model, Gemini 4 Carbon, across multiple benchmarks.
For context, in the FrontierCode 2.0 benchmark—which evaluates agent-based coding capabilities—Fable 5.2 achieved a score of 49.1% at the “Xhigh” difficulty level, whereas Gemini 4 Carbon scored 41.1% at the “medium” difficulty level. Fable 5.2 also outperformed Fable 5.1, Opus 5.5, and OpenAI’s GPT-6 Astra, which recorded scores of 32.4%, 45.4%, and 42.8%, respectively.
However, in the AutomationBench tests, Gemini 4 Carbon scored higher than Fable 5.2 in handling business workflows. The AI model in Mythos 5.2 performed better at detecting cybersecurity vulnerabilities, achieving an 89.6% CVE rediscovery rate—significantly higher than the 78.3% recorded by Mythos 5.1.
Furthermore, Anthropic claims that, at a medium difficulty level, Fable 5.2 can complete “challenging,” complex tasks using fewer tokens than Fable 5.1. Meanwhile, Mythos 5.2 is said to perform better at reproducing and patching vulnerabilities. The company notes that research has shown one-sixth of the vulnerabilities identified by Mythos 5.2 were missed by other frontier AI models.