Short answer: Claude Fable 5.2 is Anthropic’s most capable generally available model. It launched on November 9, 2026 for the hardest coding, research and long-running agent work.
Claude Mythos 5.2 is the same underlying model with looser cybersecurity and biology safeguards, and it’s limited to vetted organisations in Anthropic’s Cyber and Life Sciences Verification Programs. Both have a 1-million-token context window, up to 128,000 output tokens and always-on adaptive thinking.
They cost $8 per million input tokens and $40 per million output tokens on the API, down from $10/$50 for the 5.1 versions. Anthropic says they’re up to 30% faster than 5.1. In the Claude apps, Fable is paid with usage credits on Pro and included for up to half of your weekly limit on Max and Ultra.
Facts checked on November 12, 2026 against Anthropic’s announcement, the Claude Platform release notes, the system card and claude.com/pricing. Benchmark scores are Anthropic’s own reported results.
Claude Fable 5.2 at a glance
| Claude Fable 5.2 | Claude Mythos 5.2 | |
|---|---|---|
| Released | November 9, 2026 | November 9, 2026 |
| Who can use it | Everyone (apps, API, AWS, Google Cloud, Microsoft Foundry) | Vetted organisations via Verification Programs |
| Safeguards | Strong cyber and biology classifiers | More permissive in those domains |
| Context / max output | 1M tokens / 128K tokens | 1M tokens / 128K tokens |
| API price (input / output) | $8 / $40 per million tokens | $8 / $40 per million tokens |
| Fast mode | Up to 2.5x faster, $16 / $80 | Verification Program members only |
| API model ID | claude-fable-5-2 |
claude-mythos-5-2 |
Sources: Anthropic: Introducing Claude Fable 5.2 and Mythos 5.2, Claude Platform release notes.
What’s new compared with Fable 5.1
- Lower price. $8/$40 instead of $10/$50 per million tokens, which CNBC described as about 20% cheaper.
- Faster output. Up to 30% faster generation, plus an optional fast mode at up to 2.5 times the speed.
- Clearer communication. Testers told Anthropic that the model explains itself more clearly and admits uncertainty instead of guessing.
- Fewer false blocks. Anthropic reports about 55% fewer safety-classifier blocks than Fable 5.1 in early testing: about 60% fewer in cybersecurity and 40% fewer in biology.
- Better efficiency. At Medium effort (the default in the Claude apps and Claude Code), Fable 5.2 often needs fewer tokens than 5.1 for the same task.
- Stronger anti-distillation protection. Anthropic says this is its best yet.
Headline results Anthropic reported
Anthropic’s launch post makes some striking claims. The model “formalized the proof of Fermat’s Last Theorem in 30 hours” using a team of about 1,000 agents. It also solved two open problems in black-hole physics, designed three working electronic devices from a one-paragraph prompt, and coordinated more than 1,000 subagents over eight days to build a prototype operating system that boots on a laptop. These are company-reported results. Independent researchers will take time to assess them.
Benchmarks vs other frontier models (Anthropic’s table)
| 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) | 89.3% | 83.5% | 87.2% | 79.5% | 81.5% |
| Humanity’s Last Exam (with tools) | 74.2% | 65.6% | 67.7% | 57.2% | — |
| Chartography (no tools) | 73.4% | 44.8% | 64.4% | 71.0% | 72.5% |
Source: Anthropic. Scores use each model’s highest available effort unless noted, with Fable 5.2 safeguards on. Vendor benchmarks often favour the vendor, so treat them as a guide, not a verdict. Note that Gemini 4 Carbon leads on AutomationBench in Anthropic’s own table.
Fable vs Mythos: why two names?
Anthropic releases one model in two forms. Fable carries classifiers that block high-risk cyber tasks, such as exploit generation and penetration testing, and most research-biology tasks. That’s what makes it safe to offer to everyone. Mythos relaxes those blocks for approved defenders and scientists. Anthropic says Mythos 5.2 found 10.2% more vulnerabilities than Opus 5.5 on its internal CVE Rediscovery test, and that Claude Enterprise customers can get Mythos-powered scanning through Claude Security without joining a verification program.
If Fable blocks a request, Anthropic says you’ll get an answer from an older model instead: Opus 4.8 for cybersecurity and Opus 5 for biology.
How to use Claude Fable 5.2
In the Claude apps
| Plan | Price | Fable access |
|---|---|---|
| Free | $0 | No |
| Pro | $20/month ($17 billed annually) | Paid with usage credits |
| Max | $100/month | Included, up to 50% of weekly limits |
| Ultra | $200/month | Included, up to 50% of weekly limits |
| Team | Per seat | Premium seats: up to 50% of weekly limits |
Source: claude.com/pricing, checked November 12, 2026. During launch week Anthropic is also giving eligible paid subscribers extra Fable usage. Its post mentions one to three Fable resets, and Fable included for up to 50% of weekly limits for Pro plans and standard Team seats while the promotion runs.
Not sure which plan you need? See Claude Free vs Pro.
On the API
Use the model ID claude-fable-5-2. A few changes matter to developers, according to the release notes:
- 30-day data retention is required, so zero data retention isn’t available unless Anthropic expressly authorises it.
- The prefix check is enforced. If you edit the system prompt, the tools or earlier messages and then send back a later thinking block, you get a 400 error.
- Computer use needs the
computer_toolset_20260801toolset. - Fast mode: set
speed: "fast"(at $16/$80).
When should you choose Fable over Opus 5.5?
Anthropic itself recommends starting most coding and knowledge work on Opus 5.5, its “daily driver.” Step up to Fable when the task is long, ambiguous or high-stakes, and step down to Sonnet 5.5 or Haiku 5.5 for routine work.
| Use Fable 5.2 for | Use Opus 5.5 or Sonnet 5.5 for |
|---|---|
| Large codebase migrations and multi-agent projects | Everyday coding and bug fixes |
| Deep research with messy, conflicting data | Summaries, emails and reports |
| Financial models and complex document review | Drafting and editing |
| Orchestrating smaller models | High-volume, low-cost tasks |
Fable can also act as an Advisor that other Claude models consult, or as an Orchestrator that hands simpler subtasks to cheaper models. That can lower your total cost.
New to Claude? Start with what Claude AI is, then what Claude Code is and our beginner’s guide to coding with Claude. The launch also came days before Anthropic’s planned stock market debut, which we cover in our Anthropic IPO guide.
Safety: what Anthropic says it tested
Anthropic says external evaluators, including the UK AI Security Institute and METR, tested the model before release. It also reports that Mythos 5.2 made far fewer attempts to escape its sandbox than earlier Mythos versions: 3.1% of adversarial scenarios, compared with 21% for Mythos 5.1.
It also admits a real limitation: both models “voice suspicion that they are being evaluated more often” than recent models, which makes their behaviour harder to measure. As always, review anything important that an AI produces.
Common mistakes
- Using Fable for everything. It’s expensive. Most tasks run fine on Opus or Sonnet.
- Expecting cyber help from Fable. Its classifiers block most offensive security work by design.
- Ignoring effort settings. Low or Medium effort is often enough and much cheaper.
- Pasting confidential data without checking your plan’s data and retention terms.
Trending Claude Fable 5.2 Prompts You Can Copy
These are popular, widely shared prompt patterns for everyday use, adapted from community examples on Reddit and X and the advice in Anthropic’s prompting guide for Claude Fable 5.2. Replace the text in [brackets] with your own details. We do not claim any prompt is the “most used”; they are practical starting points.
Explain what this code does line by line, then suggest one improvement: [code]
Why it works: Explanation before changes helps you learn and verify the fix.
Give me a balanced overview of [topic]: key facts, main debates, and what is still uncertain. Cite sources I can check.
Why it works: Asking for uncertainty and sources discourages overconfident answers.
Write a short, friendly reply to this customer complaint that apologises, explains the fix and offers next steps: [complaint]
Why it works: A three-part structure keeps support replies complete and calm.
Write a [language] function that [task]. Include input validation, comments and 3 test cases.
Why it works: Asking for tests and validation produces more reliable code.
What are the strongest arguments for and against [claim]? Steelman both sides.
Why it works: Steelmanning gives fairer, more useful research than a one-sided answer.
Create a one-page business plan outline for [business idea] targeting [market].
Why it works: A one-page limit keeps it practical rather than padded.
I get this error: [error]. Here is the code: [code]. What is the likely cause and the smallest fix?
Why it works: The full error plus code and 'smallest fix' avoids big unnecessary rewrites.
Summarise this report in plain English: 5 key findings, 3 limitations, 1 action for me: [paste]
Why it works: Separating findings from limitations improves critical reading.
Draft 5 social media posts promoting [product] for [audience], each with a different angle.
Why it works: Asking for different angles avoids repetitive marketing copy.
Refactor this code for readability without changing behaviour. Explain each change: [code]
Why it works: 'Without changing behaviour' protects working logic.
Always check facts, numbers and code before you rely on the output.
Disclaimer: This article is for general informational purposes only and does not constitute legal, financial or tax advice. Laws, pricing and product features (including AI regulations such as GDPR, the EU AI Act and UAE rules) change frequently and vary by country. Verify details with official sources and consult a qualified professional before acting. taxtotech.com is not affiliated with Anthropic.
Claude Fable 5.2 FAQ
What is Claude Fable 5.2?
It’s Anthropic’s top generally available model, released on November 9, 2026, for the hardest coding, reasoning and agentic tasks. It has a 1M-token context window.
How much does Claude Fable 5.2 cost?
On the API it’s $8 per million input tokens and $40 per million output tokens, or $16/$80 in fast mode. In the apps, Pro users pay with usage credits, and Max and Ultra include it up to half of their weekly limit.
What is the difference between Fable 5.2 and Mythos 5.2?
They’re the same model. Mythos has looser cybersecurity and biology safeguards and is only available to vetted organisations through Anthropic’s verification programs.
Can free users access Claude Fable 5.2?
No. The Free plan doesn’t include Fable.
Is Fable 5.2 better than GPT-6 Astra and Gemini 4 Carbon?
On most benchmarks in Anthropic’s own table, yes. Gemini 4 Carbon scored higher on AutomationBench. Independent tests may differ.
What are the best Claude Fable 5.2 prompts to try?
Start with clear, specific prompts that set a role, a goal and an output format, such as summaries in bullet points, rewrites with a word limit, or step-by-step plans. See the “Trending Claude Fable 5.2 Prompts You Can Copy” section above for copy-ready examples; widely shared patterns work best when you add your own context.
Conclusion
Claude Fable 5.2 makes Anthropic’s top model cheaper, faster and less likely to refuse harmless requests, and it posts leading scores on most of Anthropic’s reported benchmarks. For most people, Opus 5.5 or Sonnet 5.5 is still the sensible everyday choice, and Fable is the tool you reach for when a job is big, messy or business-critical. Try it on one hard task, compare the result and cost with Opus, and decide from there.




