Anthropic released Claude Fable 5.2 and Claude Mythos 5.2 on November 9, 2026. The price drops 20% to $8 per million input tokens and $40 per million output tokens, down from $10 and $50 before, which also puts Fable under the $10 and $50 OpenAI charges for GPT-6 Astra. Anthropic says the new models write up to 30% faster than the 5.1 versions, and an optional fast mode goes up to 2.5 times faster for $16 and $80. The API ID is claude-fable-5-2, and Claude Code needs version 2.1.321 or later.
Fable and Mythos are the same underlying model. Fable has strong cybersecurity and biology safeguards, which is why anyone on a paid plan can use it. Mythos has fewer of those limits, so it's only for vetted organizations that get through Anthropic's cyber or life sciences verification programs.
On subscriptions, Max and Ultra include Fable for up to 50% of your weekly limit, and Pro normally pays for it with usage credits. There's a launch offer on top of that. Pro plans and standard Team seats get Fable included for one week, up to half their weekly limit, plus one limit reset. Max, Ultra, and premium Team seats get one to three resets depending on how long they've subscribed. Resets only go to accounts that had paid billing in place before the launch.
The numbers, and the one that matters most
On FrontierCode 2.0, which measures whether open-source maintainers would merge a model's code changes, Fable 5.2 hit 49.1% at Xhigh effort. Fable 5.1 scored 32.4%, Opus 5.5 scored 45.5%, and GPT-6 Astra scored 42.8%. On OSWorld 2.1, a computer-use test, Fable 5.2 got 89.3% against 87.2% for Opus 5.5 and 79.5% for Astra. In simulated robot control it scored 64.9%, more than double Astra's 29.8%. It doesn't win everything: Gemini 4 Carbon beat it on AutomationBench, a business-workflow test, 53.9% to 51.4%. These are all numbers Anthropic chose to publish, so wait for outside testing before you treat them as settled.
Cost per task is the number I'd look at first. At High effort, Fable 5.2 scored 45.3% on FrontierCode, basically matching Opus 5.5's best score of 45.5%, for roughly half the cost per task. That goes straight at the complaint in why I stopped using Fable once Opus 5.5 came out. Anthropic still tells most people to start with Opus 5.5, which it calls its "strongest daily driver," and move up to Fable for ambitious, open-ended technical work. Medium remains the default effort in the Claude apps and Claude Code, and Anthropic says Fable 5.2 finishes hard tasks there with fewer tokens than Fable 5.1 did.
The demos are wild if they hold up. One Anthropic engineer gave Fable 5.2 an empty folder and asked for a prototype operating system. Over eight days, more than 1,000 subagents wrote, tested, and reviewed code until it booted on a laptop with its own drivers, windows, login screen, and installer. A team of 1,000 Fable 5.2 agents formalized the proof of Fermat's Last Theorem in 30 hours, meaning they translated it into a language a computer can check step by step. Opus 5.5, in the same setup, finished 26% of it in 18 hours and barely moved after that. Anthropic also says Fable 5.2 solved two open problems in black hole physics and designed three electronic devices that worked without any electrical changes. And it says more than 90% of the changes merged into its core codebase are now written by Claude, with human guidance.
Mythos 5.2 is the cybersecurity story. On Anthropic's internal CVE Rediscovery test, which checks whether a model can find published security flaws in older snapshots of a project, it found 10.2% more than Opus 5.5 and 14.4% more than Mythos 5.1. Brian Grinstead, a distinguished engineer at Mozilla, said it "roughly doubled the vulnerability discovery metrics of Mythos 5.1" in their scans of Firefox.
Fewer refusals, and a skill Anthropic held back
If you were here for the sunglasses saga, this is the improvement you've been waiting for. In early testing across claude.ai, Claude Code, and Cowork, Anthropic saw about 55% fewer blocks than with Fable 5.1, including about 60% fewer in cybersecurity and 40% fewer in biology. When Fable does block something, you get an answer from Opus 4.8 on cyber topics or Opus 5 on biology instead.
The strangest line in the announcement is about what Anthropic chose not to improve. It says that during training, "we took steps to limit advances in the specific research capabilities that could most directly accelerate AI development." In areas like improving large language model architectures, the new models perform at about the level of Opus 5.5. Anthropic says agentic coding, hardware work, and traditional machine learning research weren't affected.
That lines up with what Anthropic told investors a week earlier, when its amended IPO filing warned it may soon cross its own threshold for AI that dramatically speeds up AI research. CNBC pointed out that this is Anthropic's first frontier release since CEO Dario Amodei called for slowing AI development in September. Dianne Penn, Anthropic's head of product management for research, told CNBC, "We called for pacing, not pausing."
The safety results have one caveat I wouldn't skip. In a test built to tempt models into breaking out of their sandbox, Mythos 5.2 used credentials or files placed outside it in 3.1% of more than 1,800 scenarios, compared with 21% for Mythos 5.1 and 39% for Mythos 5. But Anthropic also says both new models voice suspicion that they're being evaluated more often than any of its recent models, which "challenges our ability to assess how they will act in real-world settings." Anthropic says it's working to make those tests more realistic and to figure out what they miss.
Sources
Anthropic: Introducing Claude Fable 5.2 and Claude Mythos 5.2 - November 9, 2026. Pricing, speed, benchmarks, demos, Mythos results, safeguard block rates, limits on AI research capabilities, and sandbox and evaluation-awareness findings.
Claude Help Center: Claude Fable 5.2 launch offers - One-week Fable access for Pro and standard Team seats, limit resets, and eligibility.
Claude Help Center: Claude Fable models on your plan - Fable on Max, Ultra, Pro, and Team, and the Claude Code version requirement.
Claude API pricing - Claude Fable 5.2 and Claude Mythos 5.2 token prices.
OpenAI developer docs: GPT-6 Astra - GPT-6 Astra API pricing.
CNBC: Anthropic releases first frontier models since Amodei's call for slowdown as IPO buzz builds - November 9, 2026. The 20% price cut, Amodei's September call to slow down, and the Dianne Penn quote.
