Anthropic releases Claude Fable 5.2 and Mythos 5.2: prices, fast mode and access
Anthropic's Claude Fable 5.2 is out for all customers at $8/$40 per million tokens, with an optional fast mode. Mythos 5.2, the same model without safeguards, stays restricted to vetted teams.
- Author
Umair Mateen- Published
- Reading time
- 4 min
- Category
- AI
Ideas in this post
Tap one for a 5-second answer. Underlined words in the article work the same way.
0 of 4 understood
Anthropic released Claude Fable 5.2 and Claude Mythos 5.2 on 9 November 2026, upgrades to its two most capable models. In its announcement, Anthropic says the new versions improve coding, reasoning and vision. Fable 5.2 is open to all customers; Mythos 5.2 stays restricted to vetted organisations.
What was released
The two names cover one model. Anthropic puts it plainly: "Claude Fable and Claude Mythos are the same underlying model." The difference is the safety layer:
- Fable 5.2 runs with safeguards that block high-risk cyber attack work and most research-level biology. That is what lets Anthropic offer it to everyone.
- Mythos 5.2 is the same model without those safeguards. Organisations doing defensive security or biomedical work can apply through Anthropic's Cyber Verification Program or Life Sciences Verification Program. Claude Enterprise customers can also reach it through Claude Security, Anthropic's security-scanning product.
Anthropic's own recommended default for everyday work remains Claude Opus 5.5, with Sonnet 5.5 and Haiku 5.5 as cheaper options. Fable is positioned for the hardest technical jobs.
Quick check · Myth or fact
Tokens: myth or fact?
Decide for yourself first, then tap each statement to check it.
0 of 3 checked
From the explainer What are tokens in AI? Explained with word-game tiles
Pricing and fast mode
- Standard: $8 per million and $40 per million output tokens. A token is a small chunk of text, roughly three-quarters of a word.
- Fast mode (optional): up to 2.5 times faster output, at $16 input and $80 output per million tokens. Anthropic says fast mode is available in Claude Code, on the Claude Platform, on Claude on Google Cloud and in Microsoft Foundry. For Mythos, fast mode is limited to organisations in the verification programmes.
What the numbers say
Anthropic published a comparison table against Opus 5.5 and rival models. All figures below are Anthropic's own and have not been independently checked:
- Software engineering (FrontierSWE v2): Fable 5.2 scored 69.9%, against 62.3% for Opus 5.5.
- Agentic coding (FrontierCode 2.0): 49.1% at the highest effort setting, against 45.5% for Opus 5.5. At the lower "High" setting, Fable roughly matched Opus 5.5 at about half the cost per task, Anthropic says.
- Remote Labor Index: 41.3%, against 21.3% for Opus 5.5.
- Computer use (OSWorld 2.1, partial offline set): 89.3%, against 87.2% for Opus 5.5.
Not every row favours Fable: on AutomationBench, Google's Gemini 4 Carbon scored higher (53.9% against 51.4%).
On security, Anthropic says Mythos 5.2 found 10.2% more known vulnerabilities than Opus 5.5 in an internal test covering 1,829 CVEs (publicly catalogued security flaws).
Safety and testing
Anthropic says it rates the risk of the models behaving unsafely in high-stakes situations as low, in line with its recent models. External testing involved the UK AI Security Institute, METR and Frontier Design. In an automated audit of roughly 1,900 scenarios, Anthropic reports Mythos 5.2 matched or beat Opus 5.5 and Mythos 5.1 on most measures, including fewer false claims that a task was finished. Anthropic has published a full alongside the announcement.
Who should care
- Engineering teams with long, difficult coding or agent tasks where Opus 5.5 falls short.
- Teams running multi-step , which can now have a cheaper model call Fable only for the hard parts. Anthropic describes two set-ups: Fable as an "advisor" that smaller models consult, or as an "orchestrator" that hands simple subtasks to smaller models.
- Security teams at organisations that could qualify for Mythos access or already use Claude Enterprise.
Limits and caveats
- Cost. At $8/$40 per million tokens, Fable is a premium model. For routine chat, summaries and simple automations, Opus 5.5 or a smaller model will usually be the better value.
- Blocked requests. Fable's safeguards will stop some legitimate security and biology work. Teams in those fields should plan for the Mythos application route.
- Vendor . The scores come from Anthropic. Test on your own tasks before switching.
What to do this week
- Pick two or three of your hardest real tasks that currently fail or need heavy rework, and run them on Fable 5.2 at standard and lower effort settings. Compare quality and cost per task, not per token.
- If you run agents, try routing only the difficult steps to Fable and leave the rest on a cheaper model.
- Check the models overview for the exact model name and availability on your cloud before changing production settings.
- Security or life-sciences teams: review whether you qualify for a verification programme.
The DevDrift helps teams build AI automation into their workflows, with approvals and audit trails, and pick the right model for each step.
Sources
You’ve read the news. Now understand it.
- Learn · 5 min · Myth or factWhat are tokens in AI? Explained with word-game tilesTokens are the small pieces of text an AI model reads and writes, and the unit used to measure limits and cost. Explained with a word game whose tiles carry chunks of words.Start in 5 seconds →
- Learn · 5 min · Myth or factWhat is a model card? Explained with the label on a food packetA model card is the document published with an AI model that says what it is for, how it was tested and where it falls short. Think of it as the label on a food packet.Start in 5 seconds →
- Learn · 5 min · Myth or factWhat is an AI agent? Explained with an assistant you send on errandsAn AI agent takes a goal, plans the steps and carries them out with software tools, checking each result as it goes. Explained with an assistant you send on errands.Start in 5 seconds →
- Learn · 5 min · Myth or factWhat are AI benchmarks? Explained with a school examAn AI benchmark is a standard test: the same questions and the same marking for every model, so scores can be compared. Explained with a school exam that every student sits.Start in 5 seconds →
