Screenshots show the official pages on the capture date and may differ from the pages at publication. Select an image to enlarge it.
On November 9, 2026, Anthropic released two AI models, "Claude Fable 5.2" and "Claude Mythos 5.2." Both are the company's most capable models, aimed at technically demanding work such as programming and scientific research. The two are the same underlying model: Fable 5.2 is offered to the public with strong safeguards in cybersecurity and biology, while Mythos 5.2 has some of those safeguards removed and is offered only to vetted organizations.
Main changes from Fable 5.1
According to Anthropic, Fable 5.2 and Mythos 5.2 improve on the previous generation, Fable 5.1 and Mythos 5.1, in the following ways.
- Better performance in programming, reasoning and image understanding
- Up to 30% faster at generating responses
- Lower API pricing
- Fewer "unnecessary blocks," where harmless requests are mistakenly refused, while keeping safeguards just as strong. In early testing, there were about 55% fewer blocks than with Fable 5.1 (about 60% fewer in cybersecurity and about 40% fewer in biology)
Anthropic says Fable 5.2 is a step above "Claude Opus 5.5," its main model for everyday use, on almost every capability it evaluated. Even so, it recommends starting most programming and knowledge work with Opus 5.5 and switching to Fable or Mythos when needed.
Performance comparison
Here are some of the evaluation results Anthropic published (in the order Fable 5.2 / Fable 5.1 / Opus 5.5 / OpenAI's GPT-6 Astra / Google's Gemini 4 Carbon).
- Real-world software development tasks (FrontierSWE v2): 69.9% / 56.3% / 62.3% / 65.5% / 61.2%
- Difficult programming tasks (FrontierCode 2.0): 49.1% / 32.4% / 45.5% / 42.8% / 41.1%
- Computer use (part of OSWorld 2.1): 89.3% / 83.5% / 87.2% / 79.5% / 81.5%
- Hard questions across many fields (Humanity's Last Exam, with tools): 74.2% / 65.6% / 67.7% / 57.2% / not published
- Business automation (AutomationBench): 51.4% / 31.4% / 42.5% / 41.4% / 53.9%
Fable 5.2 does not come out on top in every category. On the business automation evaluation, for example, Gemini 4 Carbon scores higher. These results were also measured by Anthropic itself.
As examples of what the model can do, Anthropic says it ran about 1,000 AI agents, also drawing on propositions and lemmas prepared in advance, to rewrite the proof of "Fermat's Last Theorem" in mathematics into a form a computer can verify in 30 hours, solved two open problems in black hole physics, and rewrote a production service from Python to Rust, making it start about 26 times faster.
Pricing and where to use it
Fable 5.2 is available on Claude's Pro, Max, Ultra, Team and Enterprise plans. In the Claude apps and the developer tool Claude Code, it runs with the thinking depth set to "Medium." As a bonus in the week of release, Fable usage allowances are reset one to three times depending on how long users have been subscribed. In addition, standard seats on Pro and Team can use Fable for up to 50% of their weekly usage limit.
The API model name is claude-fable-5-2. Pricing is $8 per 1M input tokens and $40 per 1M output tokens, 20% lower than Fable 5.1's $10 and $50. The amount of text it can handle at once (its context) is 1M tokens, and a single output can be up to 128,000 tokens. It is also available through Amazon Web Services, Google Cloud and Microsoft Azure (Microsoft Foundry). A "fast mode" that generates responses up to 2.5 times faster costs $16 for input and $80 for output, and is available in Claude Code, Anthropic's developer platform, Google Cloud and Microsoft Foundry.
Mythos 5.2 is only for vetted organizations
Mythos 5.2 is the same model as Fable 5.2, with some of its cybersecurity and biology safeguards removed. Anthropic calls it its strongest model for technical cybersecurity work.
In Fable 5.2, many high-risk cyber tasks, such as penetration testing and writing exploit code, along with many biology tasks meant for research, are blocked, and a different model with safeguards answers those requests instead (when using the API, this fallback has to be configured). Mythos 5.2 can do these tasks, but in exchange it is available only to organizations that have passed review through Anthropic's "Cyber Verification Program" and "Life Sciences Verification Program." Its API pricing is the same as Fable 5.2. Claude Enterprise customers can use Mythos 5.2 through "Claude Security," a feature for finding vulnerabilities.
On safety, Anthropic says outside organizations such as the UK AI Security Institute and METR tested the models before release. It also says it has made its protections against "distillation," where one AI's outputs are used to train another AI, the strongest they have ever been.
CoAI's take
By raising performance while cutting prices by 20%, Anthropic has made Fable 5.2 easier to reach for than the previous generation. Still, it costs twice as much as Opus 5.5 ($4 input and $20 output), and Anthropic itself recommends starting with Opus 5.5. Fable 5.2 is the model to pick when the performance gap directly affects results, as in long-running development or research.
Last week, OpenAI also released "GPT-6 Cyber," a security-specific model, only to vetted organizations. Limiting who can use powerful capabilities that could also be used for attacks, while offering a version with safeguards to the general public, is becoming the common approach among the major AI companies.
