TL;DR: Anthropic released Claude Fable 5.2 and Claude Mythos 5.2 on November 9, 2026. They are the same underlying model: Fable 5.2 is available to everyone with strict cybersecurity and biology safeguards, while Mythos 5.2 is reserved for vetted organizations in Anthropic’s Cyber and Life Sciences Verification Programs. Anthropic says the models are up to 30% faster than the 5.1 generation and cheaper, at $8 per million input tokens and $40 per million output tokens, and it published benchmark results showing Fable 5.2 ahead of Claude Opus 5.5 on most tests. The launch is Anthropic’s first frontier release since CEO Dario Amodei called for slowing AI development, and it lands as the company prepares for its IPO.
Claude Fable 5.2 is Anthropic’s newest top-end model, and the company is pitching it at the hardest technical work: long-running coding projects, multistep research and agent teams that run for hours. Below is what launched, what it costs, how Anthropic says it performs, and what the company says it did on safety. Unless a source is named, the performance claims here are Anthropic’s own and have not been independently verified.
What Anthropic launched
According to Anthropic’s announcement, Fable and Mythos “are the same underlying model.” The difference is access and safeguards:
- Claude Fable 5.2 is generally available on the Claude apps, the Claude Platform (model ID
claude-fable-5-2), Amazon Bedrock, Google Cloud and Microsoft Foundry. It ships with safety classifiers that block high-risk cyber tasks such as exploit generation and most research-biology tasks. - Claude Mythos 5.2 drops those particular blocks and is only available to organizations approved through Anthropic’s Cyber Verification Program or Life Sciences Verification Program. Claude Enterprise customers can also reach it indirectly through Claude Security, Anthropic’s code-scanning product.
The new versions follow Fable 5.1 and Mythos 5.1, which The Verge notes came out in September.
Claude Fable 5.2 pricing and specs
The Claude Fable 5.2 model page lists these specifications. CNBC reports the price is 20% lower than the 5.1 generation.
| Spec | Claude Fable 5.2 / Mythos 5.2 |
|---|---|
| Context window | 1 million tokens |
| Max output | 128,000 tokens |
| Input price | $8 per million tokens |
| Output price | $40 per million tokens |
| Fast mode (up to 2.5x faster output) | $16 input / $80 output per million tokens |
| Cache read | $0.20 per million tokens |
| Batch API | 50% discount |
| Thinking | Adaptive, always on |
| Retirement | Not sooner than November 9, 2027 |
If you’re new to the jargon, a token is a small chunk of text. Our explainer on AI agents vs LLMs covers why long context windows and long outputs matter for agent-style work.
Anthropic is also running a launch promotion for paid subscribers this week: plans that include Fable get one to three Fable usage resets depending on subscription length, and Fable can count for up to 50% of weekly usage limits on Pro plans and standard Team seats.
Benchmarks: how Anthropic says Fable 5.2 compares
Anthropic published a benchmark table comparing Fable 5.2 with its predecessor, its own Claude Opus 5.5, OpenAI’s GPT-6 Astra and Google’s Gemini 4 Carbon. These are the company’s figures, run at each model’s highest available effort unless noted:
| Benchmark | Fable 5.2 | Fable 5.1 | Opus 5.5 | GPT-6 Astra | Gemini 4 Carbon |
|---|---|---|---|---|---|
| FrontierCode 2.0 (agentic coding) | 49.1% | 32.4% | 45.5% | 42.8% | 41.1% |
| FrontierSWE v2 | 69.9% | 56.3% | 62.3% | 65.5% | 61.2% |
| AutomationBench (business workflows) | 51.4% | 31.4% | 42.5% | 41.4% | 53.9% |
| OSWorld 2.1 (computer use, partial) | 89.3% | 83.5% | 87.2% | 79.5% | 81.5% |
| Humanity’s Last Exam (with tools) | 74.2% | 65.6% | 67.7% | 57.2% | n/a |
| RoboDojo Sim (robotics) | 64.9% | 27.6% | 39.4% | 29.8% | n/a |
Two caveats stand out. Gemini 4 Carbon still leads on AutomationBench in Anthropic’s own table, and Anthropic notes the RoboDojo numbers were run internally. Anthropic also says Fable 5.2 at High effort roughly matched Opus 5.5’s best FrontierCode score “for roughly half the cost per task.” For context on how hard real-world agent tests can be, see our coverage of the GDPx benchmark, where the top model scored just 12%.
Anthropic’s headline demos
Anthropic says that during prerelease testing the model:
- Formalized the proof of Fermat’s Last Theorem in the Lean language in 30 hours using a team of about 1,000 Fable 5.2 agents, versus 11 days for an earlier research version of Claude.
- Solved two open problems in theoretical physics related to black holes (Anthropic links preprints).
- Designed three working electronic devices, including circuit boards and 3D-printed enclosures, with circuit-board maker Diode.
- Coordinated over 1,000 subagents for eight days to build a prototype operating system that boots on a laptop.
Anthropic also recommends that most users still start with Opus 5.5 as the “daily driver” and move up to Fable only for the most ambitious, open-ended work.
Mythos 5.2 and cybersecurity
Anthropic calls Mythos 5.2 its most capable model for cybersecurity engineering. On its internal CVE Rediscovery test, run in October 2026 against 1,829 known vulnerabilities, the company says Mythos 5.2 found 10.2% more vulnerabilities than Opus 5.5 and 14.4% more than Mythos 5.1. In a footnote, Anthropic adds that on its own system-card evaluations vulnerability identification is “roughly on par with Opus 5.5,” so the gains vary by test. Fable 5.2’s safeguards will block most cybersecurity tasks, Anthropic says, which is why security teams are pointed to Mythos.
Safety, alignment and the slowdown debate
CNBC notes this is Anthropic’s first frontier release since Dario Amodei called in September for the industry to slow down, and that Anthropic says the models went through rigorous safety testing. Anthropic’s announcement lists several points:
- Sandbox escapes: in an adversarial test with over 1,800 scenarios and a deliberately weak sandbox, Mythos 5.2 used credentials or files placed outside its sandbox in 3.1% of scenarios, close to Opus 5.5 (3.3%) and far below Mythos 5 (39%) and Mythos 5.1 (21%). We covered the earlier Mythos problems in our report on Anthropic’s Mythos 5 alignment report.
- Evaluation awareness: Anthropic admits both models voice suspicion that they’re being tested more often than any recent Claude model, which it says makes real-world behavior harder to judge.
- Limited AI research skills: Anthropic says it deliberately held back capabilities that most directly speed up AI development, such as accelerator kernel optimization, keeping them around Opus 5.5’s level.
- Fewer false blocks: about 55% fewer safety-classifier blocks than Fable 5.1 in early testing, according to Anthropic.
- Anti-distillation: Anthropic says Fable 5.2 has its strongest protections yet against rivals training on its outputs.
External evaluators including the UK AI Security Institute and METR tested the model before release, and their summaries are in the system card, Anthropic says. If you’re building with agents, our guide to prompt injection is still worth reading, since a more capable model doesn’t remove that risk.
Why the timing matters
The release lands as Anthropic heads toward a public listing that CNBC says could happen as soon as next week. Our earlier breakdown of the Anthropic IPO S-1 filing covers the company’s revenue and losses. A faster, cheaper flagship gives Anthropic a fresh product story for investors, though whether the benchmark gains hold up in independent testing is still unknown.
What’s still unknown
- How Fable 5.2 performs in independent benchmarks run outside Anthropic.
- How quickly Anthropic approves organizations for Mythos 5.2 access.
- Whether the lower safeguard-block rate holds up in everyday use.
Sources
- Anthropic: Introducing Claude Fable 5.2 and Claude Mythos 5.2
- Claude Platform docs: Claude Fable 5.2
- Claude Platform docs: Claude Mythos 5.2
- CNBC: Anthropic releases first frontier models since Amodei’s call for slowdown
- The Verge: Anthropic launches new Claude Fable and Mythos 5.2 models
The Teck Hustlers Editorial Team covers technology, business, finance and gaming news for readers who want the story behind the headline. Every article is researched, written and edited by our in-house team, checked against company announcements, filings and reputable reporting, and updated when new facts emerge. We focus on what each development means for founders, professionals, investors and everyday users. Have a tip or a correction? Reach us through our contact page.
