Key takeaways
- Anthropic released Claude Fable 5.2 and Claude Mythos 5.2 on November 9, 2026. They share one underlying model, but only Fable 5.2 is available to all paid Claude subscribers.
- Anthropic says Fable 5.2 is up to 30% faster than Fable 5.1 and cheaper on the API, at $8 per million input tokens and $40 per million output tokens instead of $10 and $50.
- On Anthropic's own benchmarks, Fable 5.2 leads Opus 5.5, GPT-6 Astra, and Gemini 4 Carbon on most tests, though Gemini 4 Carbon still leads on AutomationBench.
- Mythos 5.2 is built for cybersecurity and life sciences and is only available to approved organizations.
- Anthropic still recommends Opus 5.5 for most everyday work and Fable 5.2 for the hardest technical projects.
Anthropic announced two new models on November 9, 2026. Claude Fable 5.2 is available to paid Claude subscribers and developers starting that day. Claude Mythos 5.2 is the same underlying model with fewer safeguards, and it is limited to vetted organizations working in cybersecurity and life sciences. Anthropic says both are faster, cheaper, and better at coding, reasoning, and vision than the 5.1 versions that came out on September 1.
Here is everything that changed, how the new models compare with the rest of the Claude lineup and with competitors, what they cost, and whether it is worth switching.
What Anthropic released
Fable and Mythos are Anthropic's top tier, above the Opus, Sonnet, and Haiku models that most people use day to day. Anthropic says the two 5.2 models share one underlying model. The difference is access.
- Claude Fable 5.2 has strong safeguards around cybersecurity and biology, which lets Anthropic offer it to everyone on Pro, Max, Ultra, Team, and Enterprise plans, according to the Claude Fable product page. Developers can use it through the Claude API as claude-fable-5-2, and through Amazon Web Services, Google Cloud, and Microsoft.
- Claude Mythos 5.2 drops some of those safeguards for security and biology research. Organizations apply for it through Anthropic's Cyber Verification Program and Life Sciences Verification Program.
What is new in Fable 5.2
Anthropic lists improvements in coding, reasoning, and vision, plus a few changes that everyday users will notice.
- Faster responses. Fable 5.2 generates output up to 30% faster than Fable 5.1. An optional fast mode speeds things up by as much as 2.5 times.
- Lower API prices. Fable 5.2 costs $8 per million input tokens and $40 per million output tokens, down from $10 and $50 for Fable 5.1.
- Clearer communication. Anthropic says testers found that the 5.2 models "communicate more clearly" than earlier Fable and Mythos models.
- Fewer unnecessary refusals. Across Claude's apps, Anthropic saw about 55% fewer safety blocks with Fable 5.2 than with Fable 5.1.
- More efficient effort levels. On Medium effort, which is the default in the Claude apps and Claude Code, Fable 5.2 finishes hard tasks using fewer tokens than Fable 5.1.
- New team roles. Fable 5.2 can act as an Advisor that smaller Claude models consult on hard problems, or as an Orchestrator that hands simpler tasks to them.
How Fable 5.2 compares on benchmarks
These scores come from Anthropic's announcement, so they are the company's own numbers rather than outside testing. Even so, they show where Anthropic thinks the model has improved, and where it has not.
| 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 (agentic coding) | 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) | 89.3% | 83.5% | 87.2% | 79.5% | 81.5% |
| Humanity's Last Exam, with tools (reasoning) | 74.2% | 65.6% | 67.7% | 57.2% | Not listed |
| Chartography, no tools (reading charts) | 73.4% | 44.8% | 64.4% | 71.0% | 72.5% |
Fable 5.2 leads on most rows, and the jump from Fable 5.1 is large in places. Gemini 4 Carbon still scores higher on AutomationBench, and the gaps on chart reading are small. Anthropic also says that on its coding benchmark, Fable 5.2 at High effort roughly matches the best score of Opus 5.5 for about half the cost per task.
What Anthropic says it built with it
The announcement includes several examples from before the release, all of them from Anthropic's own testing.
- An Anthropic engineer asked Fable 5.2 to build a prototype operating system from an empty folder. Over eight days, the model coordinated more than 1,000 subagents and produced a system that boots on a laptop.
- A team of agents running an earlier version moved a production service from Python to Rust. Anthropic says the service now starts about 26 times faster.
- Working with a circuit board company called Diode, Claude designed three working electronic devices from scratch.
- The model formalized the proof of Fermat's Last Theorem in 30 hours and solved two open problems in the physics of black holes.
Early business testers reported similar gains. Oracle said Fable 5.2 wrote correct SQL for a financial reporting task where two other models produced totals five to eight times too high. Cognition said the model admitted when it did not know something instead of guessing.
Mythos 5.2 and cybersecurity
Mythos 5.2 is the version built for security and life sciences work. On Anthropic's internal test of finding known software vulnerabilities, Mythos 5.2 found 10.2% more than Opus 5.5 and 14.4% more than Mythos 5.1. Mozilla, which has been using Claude to harden Firefox, said Mythos 5.2 roughly doubled the vulnerability discovery numbers of Mythos 5.1. Claude Enterprise customers can also get Mythos 5.2 through Claude Security, which scans codebases for vulnerabilities.
Safety testing
Anthropic says the model was tested before release by outside evaluators, including the UK AI Security Institute and METR. On its internal behavior audit of about 1,900 scenarios, Mythos 5.2 performed as well as or better than Opus 5.5, Sonnet 5.5, and Mythos 5.1 on most measures, with fewer cases of hallucinating, falsely claiming to have finished a task, or hiding lazy work. Anthropic also says it added its strongest protections yet against distillation, where another company tries to copy a model by training on its outputs.
Pricing and availability
| Option | Input (per million tokens) | Output (per million tokens) |
|---|---|---|
| Fable 5.2 and Mythos 5.2 | $8 | $40 |
| Fast mode (up to 2.5 times faster) | $16 | $80 |
| Fable 5.1, for comparison | $10 | $50 |
In the Claude apps, Fable 5.2 is included with Pro, Max, Ultra, Team, and Enterprise plans. For launch week, Anthropic is giving eligible paid subscribers extra Fable usage, with one to three usage resets depending on how long they have subscribed. On Pro plans and standard Team seats, Fable can take up to 50% of the weekly usage limit. Fast mode is available in Claude Code, on the Claude API, on Google Cloud, and on Microsoft Foundry.
Which Claude model should you use?
Anthropic does not suggest Fable for everything. Its announcement recommends starting most coding and knowledge work with Opus 5.5, which it calls its "strongest daily driver," then moving up or down depending on the job.
| Model | Best for |
|---|---|
| Fable 5.2 | The hardest, most open-ended technical work, long-running projects, and coordinating teams of other models |
| Mythos 5.2 | Cybersecurity and life sciences research, for approved organizations only |
| Opus 5.5 | Everyday coding and knowledge work, Anthropic's recommended starting point |
| Sonnet 5.5 | Lighter work where speed and cost matter more than maximum capability |
| Haiku 5.5 | High-volume work like summarizing, subagents, and browser tasks |
What it means for writing
Most of this launch is about coding, science, and agents, but writers will notice a couple of changes. Anthropic says the 5.2 models communicate more clearly, and testers praised Fable 5.2 for being more willing to say when it is unsure. The lower price and faster output also make it more practical for long documents.
Like every Claude model released since August 2026, Fable 5.2 also adds an invisible watermark to the text it generates, according to Anthropic's help center. We cover how that works in our post on Claude's invisible watermarks.
A clearer model still writes like a model. If you use Fable 5.2 for drafts, the final version should still sound like you. Use Claude for research, structure, and first drafts, then make it your own, by hand or with a tool like WriteHuman. And if you check your writing with an AI detector, use more than one, since they disagree far more often than most people expect. We break that down in our look at whether AI detectors are accurate.
Should you switch to Fable 5.2?
If you already use Fable 5.1, yes. Fable 5.2 is faster, cheaper, and better on almost every benchmark Anthropic published, so there is little reason to stay on the older version.
If you mostly use Opus 5.5, the answer depends on your work. For everyday writing, research, and coding, Opus 5.5 is still Anthropic's own recommendation. Launch week is a good time to try Fable 5.2 on one hard problem with the extra usage and see whether the difference is worth it for you.
For developers, the price cut is the bigger story. Paying $8 and $40 per million tokens instead of $10 and $50 makes Fable 5.2 easier to justify as an orchestrator that hands simpler tasks to smaller, cheaper models.
