Claude Fable 5.2
- Provider
- Anthropic
- Status
- Current
- Context
- 1,000,000 tok
- Price
- $8 / $40 /MTok
- Knowledge
- 2026-06
Claude Fable 5.2 is Anthropic’s Mythos-class model, released on 9 November 2026 as claude-fable-5-2 on the Claude API, in the Claude apps, and on Amazon Bedrock, Google Cloud and Microsoft Foundry (Anthropic). It succeeds Claude Fable 5.1 at $8/$40 per million tokens, a fifth below Fable 5.1’s $10/$50. Its restricted twin, Claude Mythos 5.2, is the same model with some cyber and biology safeguards removed.
Artificial Analysis scored it 65 on its Intelligence Index v5.0.1 at max effort on launch day, the highest score it had recorded on v5: five points above Gemini 4 Carbon (60) and six above Claude Opus 5.5 (59). It is also the most expensive model to run on that index, at $26.54 per completed task, because it uses a great many output tokens. Anthropic published no SWE-bench figure.
Quick specs
| Provider | Anthropic |
| Released | 9 November 2026 (Claude API, Claude apps, Bedrock, Google Cloud, Microsoft Foundry) |
| API model ID | claude-fable-5-2 |
| Context window | 1,000,000 tokens |
| Max output | 128,000 tokens |
| Knowledge cutoff | June 2026 |
| Input price | $8.00 / MTok |
| Output price | $40.00 / MTok |
| Cache reads | $0.20 / MTok |
| Fast mode | $16 / $80 per MTok |
| Default effort | High on the API; medium in the Claude apps and Claude Code |
| Modalities | Text and image in; text out |
| AA Intelligence Index | 65 at max effort at launch (v5.0.1); 62 xhigh, 62 high, 56 medium, 54 low |
| Cost per AA index task | $26.54 at max effort; $9.08 at high |
| Limitations | No SWE-bench figure; slow and token-hungry at max effort; blocked cyber and biology requests re-route to older models |
What’s new in Fable 5.2
The price. $8/$40 against Fable 5.1’s $10/$50, with cache reads at $0.20 rather than $0.25. Batch requests are half price.
The speed. Anthropic says outputs come up to 30% faster than Fable 5.1, and a new fast mode, at $16/$80, generates tokens up to 2.5 times faster.
Fewer blocked requests. Anthropic says early testing across claude.ai, Claude Code and Cowork showed about 55% fewer safeguard blocks than Fable 5.1: about 60% fewer in cybersecurity and 40% fewer in biology. Requests that are still blocked are answered by Claude Opus 4.8 for cyber and Claude Opus 5 for biology.
Fermat’s Last Theorem. Anthropic says a team of 1,000 Fable 5.2 agents fully formalised the proof of Fermat’s Last Theorem in Lean in 30 hours, with minimal human help. In the same setup, Opus 5.5 completed 26% of the formalisation in 18 hours. Anthropic’s earlier formalisation, with an unreleased research model and a purpose-built harness, took 11 days.
Anthropic still recommends Opus 5.5 as the starting point for most coding and knowledge work, with Fable 5.2 for the hardest tasks.
Benchmark performance
Anthropic’s launch table
| Benchmark | Fable 5.2 | Fable 5.1 | Opus 5.5 | GPT-6 Astra | Gemini 4 Carbon |
|---|---|---|---|---|---|
| FrontierCode 2.0 | 49.1% | 32.4% | 45.5% | 42.8% | 41.1% |
| FrontierSWE v2 | 69.9% | 56.3% | 62.3% | 65.5% | 61.2% |
| AutomationBench | 51.4% | 31.4% | 42.5% | 41.4% | 53.9% |
| Terminal-Bench-Science 0.1 | 74.6% | 52.6% | 58.7% | 68.1% | 64.8% |
| Humanity’s Last Exam (tools) | 74.2% | 65.6% | 67.7% | 57.2% | — |
| OSWorld 2.1 (partial) | 89.3% | 83.5% | 87.2% | 79.5% | 81.5% |
| Remote Labor Index | 41.3% | 17.9% | 21.3% | 20.8% | — |
Figures are at the highest effort each model offers, except FrontierCode 2.0 (Fable 5.2 and 5.1 at xhigh, Gemini 4 Carbon at medium) and Remote Labor Index (Fable 5.2 at high). These are vendor figures, run by Anthropic. At high effort Fable 5.2 scores 45.3% on FrontierCode 2.0, about level with Opus 5.5’s best, at roughly half the cost per task, Anthropic says.
The independent read (Artificial Analysis)
| Measure | Claude Fable 5.2 at launch (9 Nov 2026) | Context |
|---|---|---|
| Intelligence Index v5.0.1 | 65 (max effort) | Gemini 4 Carbon 60, Claude Opus 5.5 59, GPT-6 Astra 57 |
| Index by effort | 54 low, 56 medium, 62 high, 62 xhigh, 65 max | API default is high |
| Cost per index task | $26.54 (max); $14.83 (xhigh); $9.08 (high) | Opus 5.5 at max: $10.04 |
| Output tokens, whole index | About 275 million | |
| Output speed | 98 tokens per second | |
| Time to first answer token | About 220 seconds |
Source: Artificial Analysis, measured with Anthropic’s default fallback enabled. At high effort it scores 62, still above every other model on the index at launch, for $9.08 a task.
Pricing breakdown
| Claude Fable 5.2 | Claude Fable 5.1 | Claude Opus 5.5 | |
|---|---|---|---|
| Input | $8.00 | $10.00 | $4.00 |
| Output | $40.00 | $50.00 | $20.00 |
| Cache reads | $0.20 | $0.25 | $0.20 |
| Batch (in / out) | $4 / $20 | $5 / $25 | $2 / $10 |
Token prices are double Opus 5.5’s. Effort decides the bill more than the rate card does: on Artificial Analysis’s measurements, a completed task costs nearly three times as much at max effort ($26.54) as at high ($9.08).
How Fable 5.2 compares
vs Claude Opus 5.5
When Fable 5.2 launched, Opus 5.5 scored 59 on the index at max effort for $10.04 a task, against Fable 5.2’s 65 for $26.54. At high effort Fable 5.2 scored 62 for $9.08, ahead of Opus 5.5 at max on both score and cost. Fable 5.2 leads Opus 5.5 on every labelled row of Anthropic’s launch table, at double the token price.
vs Gemini 4 Carbon
Gemini 4 Carbon scored 60 at high reasoning for $7.76 a task when Fable 5.2 launched, and was limited to vetted cyber defenders. Fable 5.2 leads it on Anthropic’s table except on AutomationBench.
vs Claude Fable 5.1
On the same index, Fable 5.1 read 49 at max effort when Fable 5.2 launched. Fable 5.2 is a fifth cheaper per token, Anthropic says it is up to 30% faster, and it blocks about 55% fewer requests.
Known limitations
No SWE-bench figure. Anthropic published no SWE-bench Pro or Verified result, as with Fable 5.1 and Opus 5.5.
Cost and speed at max effort. $26.54 per completed index task, 98 output tokens per second and about 220 seconds to the first answer token. The headline score is the most expensive configuration to run.
Re-routing on sensitive work. Blocked cyber requests are answered by Opus 4.8 and blocked biology requests by Opus 5, unless you are in Anthropic’s Cyber or Life Sciences Verification Program.
Plan access. Not on the Free plan, and on paid plans Fable usage is capped at a share of the weekly limit.
Frequently asked questions
When was Claude Fable 5.2 released?
9 November 2026, as claude-fable-5-2 on the Claude API, in the Claude apps and on Amazon Bedrock, Google Cloud and Microsoft Foundry.
How much does Claude Fable 5.2 cost?
$8 per million input tokens and $40 per million output tokens, a fifth below Fable 5.1. Cache reads are $0.20 per million, batch requests are half price, and fast mode costs $16/$80. On Artificial Analysis’s launch measurements a completed Intelligence Index task cost $26.54 at max effort and $9.08 at high effort.
Is Claude Fable 5.2 the best AI model?
At its 9 November launch it scored 65 on the Artificial Analysis Intelligence Index v5.0.1 at max effort, the highest score the index had recorded, ahead of Gemini 4 Carbon (60) and Claude Opus 5.5 (59). It was also the most expensive model to run on that index. For current standings, see best AI models.
What is the difference between Fable 5.2 and Mythos 5.2?
They are the same underlying model. Fable 5.2 has cyber and biology safeguards and is available to anyone on a paid plan or the API. Mythos 5.2 drops some of those safeguards and is only for organisations admitted to Anthropic’s Cyber Verification Program or Life Sciences Verification Program.
Did Claude Fable 5.2 prove Fermat’s Last Theorem?
It formalised an existing proof rather than finding a new one. Anthropic says 1,000 Fable 5.2 agents wrote the whole proof out in Lean, a language in which a computer checks every step, in 30 hours. Opus 5.5 got 26% of the way in the same setup.
What is Claude Fable 5.2’s context window?
1,000,000 tokens, with text and image input and text output, a 128,000-token maximum output and a June 2026 knowledge cutoff. The full window is billed at the standard rate.