Anthropic releases Fable 5.2 and Mythos 5.2

Anthropic's new top model costs $8 per million input tokens, adds an up to 2.5x faster mode and keeps the unblocked Mythos version for vetted groups.
What it means for founders
- Price the switch by task cost. At twice the Opus 5.5 rate, Fable needs to finish jobs with fewer tokens or retries to save money. Run your own evals on Claude at matched effort levels before moving traffic.
- Check your integration first. The model page says adaptive thinking stays on, forced tool use returns a 400 error, and sampling settings like `temperature` should be left out. Preserved thinking now covers every account, so code that edits earlier turns needs updating.
- Plan for data retention. The docs say Anthropic keeps Fable 5.2 traffic for 30 days unless it authorizes an exception. Zero data retention runs through Enterprise Frontier Safeguards, a private beta with customers like Stripe; Anthropic says eligible customers can keep it meanwhile.
- Security and biotech startups face a gate. On Fable, most of that work falls back to older Opus models. Unblocked Mythos needs verification approval, though Claude Enterprise customers can reach it through Claude Security.
The story
Anthropic launched Claude Fable 5.2 alongside Mythos 5.2 on Monday, November 9, 2026, and lowered the price of both. Per the company's announcement, both names cover one underlying model: Fable carries cyber and biology safeguards and is open to everyone, while Mythos drops those blocks for vetted organizations. As TechCrunch notes, it is Anthropic's first release since METR estimated its AI progress had roughly doubled in pace.
Price, limits and access
Fable 5.2 costs $8 input and $40 output per million tokens; TechCrunch puts the old Fable input rate at $10. The model page lists a 1M token context window, 128K tokens of maximum output, a June 2026 knowledge cutoff and `high` as the default effort. Cache reads cost $0.20 per million tokens, and batch jobs get 50% off, per the pricing docs.
Fast mode runs up to 2.5 times quicker at double the rate, $16 and $80. In the Claude apps and Claude Code, the default effort is Medium. The model ID is `claude-fable-5-2`; AWS, Google Cloud and Azure also host it. Mythos 5.2 has the same specs and prices, gated behind the Cyber and Life Sciences Verification Programs.
What Anthropic claims
Anthropic's own table puts Fable 5.2 at 69.9% on FrontierSWE v2, against 62.3% for Opus 5.5 and 65.5% for GPT-6 Astra. Anthropic says a High effort Fable 5.2 roughly matched a Max effort Opus 5.5 on FrontierCode 2.0 at about half the spend per task. It still tells users to start most tasks on Opus 5.5, at $4 and $20, and step up only for the hardest work.
Safety framing
Fable's classifiers block high-risk cyber work such as exploit generation, plus most research biology, and hand those queries to Opus 4.8 (cyber) or Opus 5 (biology). Anthropic reports about 55% fewer blocks than Fable 5.1 in early testing. It says it deliberately limited gains in skills that could speed up AI development, such as kernel optimization, which sit near Opus 5.5 level. It rates high-stakes misalignment risk as low, while admitting both models "voice suspicion that they are being evaluated more often" compared with its recent ones.
METR's November 5 assessment estimated Anthropic's gains on well-specified tasks run at about twice the earlier rate, possibly 1.3x to 3x. It said Anthropic redacted most of the evidence it found most informative, which Anthropic says was confidential material that could help rivals.
What we don't know yet
- Independent benchmark results; every score above comes from Anthropic.
- When Enterprise Frontier Safeguards leaves private beta.
- How often the classifiers still block legitimate security or biotech work.
Sources
Primary sources
Reporting
Tools in this story
We may earn a commission if you sign up through our links. It never affects our ratings or which stories we cover.
Anthropic's assistant for writing, analysis, and code
Anthropic's coding agent, from terminal to IDE to browser
More in Models & Labs
- Google releases Gemini 4 Hydrogen, its new fast modelGemini 4 Hydrogen is Google's fast, cheaper model: $1.50 per million input tokens, a 1M context, and now inside Search.Google DeepMind · 6d ago

- Google releases Gemini 4 Carbon, previews /swarm

For founders: Budget for the December price. Cost models built on the $2 and $10 rate break after November 30.
Google DeepMind · 8d ago - Meta launches Muse Ember for high-volume agent work

For founders: Price your agent by the task. Ember's advantage depends on short outputs and cheap cache hits, so a pipeline that reuses a long system prompt gains the most.
TechCrunch · 9d ago - OpenAI's GPT-6.1 Terra halves per-token API rates

For founders: Rerun your agent costs. At $1 in and $5 out per million tokens, high-volume chores like invoice entry and form filling get cheaper.
OpenAI · 23d ago - Block unveils Dessa, voice AI trained only on speech

For founders: A simpler voice stack may be coming. Many voice products today chain transcription, a language model and text to speech from separate vendors such as…
The Verge · 25d ago