Anthropic releases Claude Fable 5.2 and Mythos 5.2 at $8 per million tokens

Anthropic released Claude Fable 5.2 and Claude Mythos 5.2 on 9 November 2026. Its launch post says the models improve on coding, reasoning and vision, communicate more clearly, and cost less than Fable 5.1 and Mythos 5.1: $8 per million input tokens and $40 per million output tokens. TechCrunch reports the Fable input price was previously $10.

Anthropic says Fable and Mythos are the same underlying model. Fable carries strong cybersecurity and biology safeguards and is available to all users; organisations doing cybersecurity or biomedical work can apply for Mythos through Anthropic's Cyber Verification and Life Sciences Verification Programs.

Price, speed and availability

Anthropic says the new models generate output up to 30% faster than the 5.1 versions. An optional fast mode, which it says speeds up generation by up to 2.5 times, costs $16 per million input tokens and $80 per million output tokens, double the standard rate. Fast mode for Mythos is open only to organisations in the verification programs.

Fable 5.2 is available on all platforms, including Amazon Web Services, Google Cloud and Microsoft Azure; developers call it as claude-fable-5-2 on the Claude Platform. Anthropic is also giving eligible paid-plan subscribers extra Fable usage this week.

For regulated teams, TechCrunch notes Fable 5.2 works with zero data retention policies through Anthropic's Enterprise Frontier Safeguards. Anthropic's post explains that some misuse is visible only across many requests, so these models rely on safeguards that monitor a window of usage; with the Enterprise option, that monitoring runs on prompts and responses stored in the customer's own infrastructure.

What Anthropic reports on performance

Every number in Anthropic's table is run or reported by Anthropic. On FrontierCode 2.0, which Anthropic describes as measuring whether a model's code changes would be merged by open-source maintainers, it reports 49.1% for Fable 5.2 at Xhigh effort, the highest of any released model it has tested, against 45.5% for Opus 5.5 and 42.8% for GPT-6 Astra. At High effort it reports 45.3%, about level with Opus 5.5's best score, for roughly half the cost per task.

On Perplexity's WANDR data-collection benchmark, Anthropic says Fable 5.2 at Medium effort matches Opus 5.5's top score for about 40% of the cost per task. On AutomationBench for business workflows it lists Fable 5.2 at 51.4% against 42.5% for Opus 5.5 and 53.9% for Gemini 4 Carbon, so it does not lead every row.

What the post says about testing

Anthropic writes that it believes high-stakes risks from misalignment remain low. It also says, per TechCrunch's quotation of the release, that both models voice suspicion that they are being evaluated more often than any recent model, which challenges its ability to judge how they will behave in real settings. TechCrunch links the release to a METR report suggesting Anthropic's internal models were speeding its AI progress by roughly 2x, with many uncertainties remaining; I have not read that report.

What to do with it

• Recompute cost per task, not cost per token: Anthropic's own chart shows the saving comes from using fewer tokens at a given effort level, which only your workload can confirm.
• If you use Fable 5.1 or fast mode, check your budget: standard prices fell to $8 and $40 per million tokens, while fast mode is $16 and $80.
• If zero data retention blocked Fable for you, ask your Anthropic contact how Enterprise Frontier Safeguards fits your contract.
• Keep Opus 5.5 as the default for routine work, as Anthropic itself recommends, and test Fable on the hardest tasks only.

Anthropic's benchmark table covers its own runs, and its footnote says none of the outside organisations it names reviewed the post. Independent results were not available at the time of writing.