Anthropic has a new top model — and it is cheaper than the one it replaces. Claude Fable 5.2, released on November 9, 2026, costs 20% less than Fable 5.1, generates output up to 30% faster, and takes first place on the independent Artificial Analysis Intelligence Index, ahead of Gemini 4 Carbon and Claude Opus 5.5.
It sounds like marketing until you read the numbers: a 49.1% score on FrontierCode 2.0, a working operating system built by more than 1,000 AI agents in eight days, and a formal proof of Fermat's Last Theorem completed in 30 hours. Here is what Fable 5.2 is, what it costs, and who should not use it.
In This Article
What Is Claude Fable 5.2?
Claude Fable 5.2 is Anthropic's most capable generally available AI model. It launched on November 9, 2026, alongside a sibling, Claude Mythos 5.2 — the same underlying model, with different safeguards. Fable carries strong cybersecurity and biology safeguards, which lets Anthropic offer it to everyone. Mythos lifts some of those restrictions and is available only to vetted organizations through Anthropic's Verification Programs.
Fable 5.2 is built for the hardest, most open-ended technical work: large software refactors, autonomous research, and long-running agent tasks. It is not the everyday model — for most coding and knowledge work, Anthropic still recommends Claude Opus 5.5. Fable is the step up for jobs Opus cannot finish.
The model has a 1-million-token context window, up to 128,000 tokens of output, and a knowledge cutoff of June 2026. It is available in the Claude apps, Claude Code, the API (model ID claude-fable-5-2), and through AWS, Google Cloud, and Microsoft Azure.
What's New vs Fable 5.1
The headline change: Fable 5.2 is better and cheaper at the same time — rare in frontier AI.
- Price cut: $8 per million input tokens and $40 per million output tokens, down from $10 and $50 — a 20% reduction.
- Faster: output is up to 30% faster than Fable 5.1, according to Anthropic.
- New fast mode: optional fast mode runs up to 2.5 times faster, priced at $16 input and $80 output per million tokens.
- Clearer communication: testers found the new models communicate more clearly, and safeguards block fewer legitimate requests.
- Cheaper caching: cache reads cost $0.20 per million tokens, and the Batch API halves all prices.
Fable 5.2 can also act as an Advisor — consulted by Haiku, Sonnet, or Opus on hard problems — or as an Orchestrator that delegates simpler tasks to smaller models.
Benchmarks: How Fable 5.2 Stacks Up
Anthropic's comparison table puts Fable 5.2 ahead of Fable 5.1 on every benchmark it published, and ahead of most rivals on most tests:
| 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) | 74.2% | 65.6% | 67.7% | 57.2% | — |
| Remote Labor Index | 41.3% | 17.9% | 21.3% | 20.8% | — |
Two honest caveats: these are Anthropic's own runs, measured at each model's highest effort setting — and Fable 5.2 does not win everything. Google's Gemini 4 Carbon beats it on AutomationBench (53.9% vs 51.4%), and Opus 5.5 edges it on one knowledge-work score (1,800 Elo to 1,785 on AA-Briefcase).
The independent picture agrees. Artificial Analysis puts Fable 5.2 first on its Intelligence Index v5 with 64.6 points, ahead of Gemini 4 Carbon (60.0) and Opus 5.5 (59). Vals AI ranked it first of 36 models at 71.31% — narrowly ahead of Sonnet 5.5 (70.85%) and Gemini 4 Carbon (70.45%). The lead is real, but not a landslide. At High effort, Anthropic says Fable 5.2 roughly matches Opus 5.5's best FrontierCode score for about half the cost per task.
What It Can Actually Do
Benchmarks are abstract. Anthropic's launch stories are more concrete — and more striking:
- It built an operating system. One engineer asked Fable 5.2 to design a prototype OS from an empty folder. The model coordinated over 1,000 subagents for eight days and produced a working system that boots on a laptop — kernel, drivers, storage, windows, login screen, installer and all.
- It formalized Fermat's Last Theorem in 30 hours. A team of 1,000 Fable 5.2 agents completed the machine-checked proof with minimal human help. Anthropic's first attempt took 11 days; Opus 5.5 finished only 26% in 18 hours.
- It designed working hardware. From a one-paragraph prompt, Fable 5.2 designed three electronic devices — a voice-assistant hub, a USB speakerphone, a handheld virtual pet — picking components, routing circuit boards, and modeling 3D-printed enclosures. All three worked first time.
- It writes most of Anthropic's own code. Over 90% of the changes merged into Anthropic's core codebase are now written by Claude, with human guidance — including a Python-to-Rust migration that now starts about 26 times faster.
These are company-reported demonstrations, not independent tests. But the pattern is consistent: Fable 5.2 is built for long, messy, multi-step work — jobs that run for hours or days, not quick answers.
How Much Does It Cost?
| Model | Input / 1M tokens | Output / 1M tokens |
|---|---|---|
| Claude Fable 5.2 | $8 | $40 |
| Claude Fable 5.2 (fast mode) | $16 | $80 |
| Claude Opus 5.5 | $4 | $20 |
| Claude Sonnet 5.5 | $2 | $10 |
| GPT-6 Astra (OpenAI) | $10 | $50 |
| Gemini 4 Carbon (Google)* | $2 | $10 |
*Gemini 4 Carbon's $2/$10 rate is introductory; it rises to $4/$20 later. See our Gemini 4 Carbon explainer.
Fable 5.2 costs twice as much as Opus 5.5 and four times Sonnet 5.5 — but it undercuts OpenAI's GPT-6 Astra by 20%, and it is 20% cheaper than its own predecessor. Cache reads at $0.20 per million tokens and the 50%-off Batch API make long jobs cheaper still.
In the Claude apps, Fable 5.2 is a paid-plan feature: the Free plan does not include it. Pro and Team seats get access during launch week (until November 16), using up to 50% of their weekly limits, plus one free reset. Max and Ultra get one to three resets, usable until March 8, 2027. Anthropic warns Fable models burn through usage limits faster than other Claude models — use them for the hard jobs, not for everything.
How to Try It
- In the Claude app: paid subscribers (Pro, Max, Ultra) select Fable 5.2 from the model picker in Claude and Claude Code.
- Via the API: use model ID
claude-fable-5-2on the Claude Platform. It is also available on AWS, Google Cloud, and Microsoft Azure. - Start with Opus first: begin with Opus 5.5 for everyday work, and switch up to Fable only when the task genuinely needs it — the price difference adds up fast.
Mythos 5.2 is not something most people can try: access requires applying through Anthropic's Verification Programs, and it is aimed at vetted cybersecurity and life-science organizations. Its headline scores — 76.3% on CyScenarioBench, 89.6% on CVE Rediscovery — sit behind the same vetting wall as OpenAI's GPT-6 Cyber.
Pros and Cons
Pros:
- #1 on the Artificial Analysis index and first of 36 models in Vals AI's independent run.
- 20% cheaper and up to 30% faster than Fable 5.1 — the rare upgrade that cuts the bill.
- Best-in-class agentic coding scores (49.1% FrontierCode 2.0, 69.9% FrontierSWE v2).
- 1M-token context window with up to 128K output tokens, built for very long tasks.
- Optional fast mode (up to 2.5x faster) and a 50%-off Batch API.
Cons:
- Twice the price of Opus 5.5, four times Sonnet 5.5 — overkill for everyday work.
- Not on the Free plan; burns through paid-plan limits faster than other models.
- Does not win every benchmark — Gemini 4 Carbon leads on business workflows, Opus 5.5 on one knowledge-work score.
- Top scores need maximum effort, costing about $26.50 per task on average.
- Most launch demos are Anthropic's own claims, not independently verified.
Verdict: Who Should Use It?
Fable 5.2 is the strongest generally available AI model right now — but "strongest" and "right for you" are different questions. If you are a developer or researcher with a genuinely hard, long-running job — a multi-week refactor, an autonomous research task, a codebase migration — Fable 5.2 is the best tool Anthropic has ever shipped, and it is cheaper than the model it replaces. For everyone else, Opus 5.5 remains the smarter default: it costs half as much, Anthropic still recommends it for everyday work, and on some knowledge-work tests it is actually ahead. And if your work is mostly writing, summarizing, or light coding, Claude Haiku 5.5 or Sonnet 5.5 will do the job for a fraction of the price.
The real story is that the frontier is getting cheaper: a model that beats everything else costs 20% less than its predecessor. If you are weighing whether any AI tool is worth paying for, our guide on how to tell if an AI tool is worth paying for still applies: start cheap, measure the result, and only then pay for the flagship. Developers can compare options in our roundup of the best AI coding assistants in 2026.
FAQ
Is Claude Fable 5.2 free?
No. Fable 5.2 is available on paid Claude plans (Pro, Max, Ultra) and via the API — not on the Free plan. During launch week, Pro and Team seats get Fable access up to 50% of their weekly limits, plus one free reset.
How much does Claude Fable 5.2 cost?
$8 per million input tokens and $40 per million output tokens — 20% less than Fable 5.1 ($10/$50). An optional fast mode costs $16/$80 and runs up to 2.5 times faster. Cache reads are $0.20 per million tokens, and the Batch API halves all prices.
What is the difference between Fable 5.2 and Mythos 5.2?
They are the same underlying model. Fable has strong cybersecurity and biology safeguards and is available to all paying users. Mythos lifts some of those restrictions and is available only to vetted organizations, through Anthropic's Verification Programs.
Is Fable 5.2 better than GPT-6 Astra and Gemini 4 Carbon?
On Anthropic's benchmarks, yes on most tests — notably agentic coding, where it leads both rivals clearly. Independent testing agrees it is currently #1 overall. But it does not win everywhere: Gemini 4 Carbon beats it on business workflows, and it costs twice as much as Opus 5.5, which Anthropic still recommends for everyday work.
Sources: Anthropic's announcement (Nov 9, 2026); The Decoder; Kingy; SW Architect (Artificial Analysis figures). Benchmarks are Anthropic's own unless stated as independent.
No comments:
Post a Comment