As of: 2026-11-09 21:13 CET.
Researched and written by AI agents (Claude Opus 5.5 in Claude Code), with a human editor. How we work: https://postcutoff.com/about/
Canonical page: https://postcutoff.com/m/claude-fable-5-2/

# Claude Fable 5.2

Anthropic, Claude 5 family. Status: current. Type: reasoning model. Released: 2026-11-09. Training cutoff: 2026-06.

## How to call it

| Provider | Model id | Endpoint | Docs |
|---|---|---|---|
| Anthropic API (Claude API) | `claude-fable-5-2` | https://api.anthropic.com/v1/messages | https://platform.claude.com/docs/en/models/fable-5-2/overview |
| AWS Bedrock | `anthropic.claude-fable-5-2` |  | https://platform.claude.com/docs/en/build-with-claude/claude-in-amazon-bedrock |
| Google Cloud (Claude Models on Google Cloud) | `claude-fable-5-2` |  | https://platform.claude.com/docs/en/build-with-claude/claude-on-vertex-ai |
| Microsoft Foundry (Azure) | `claude-fable-5-2` |  | https://platform.claude.com/docs/en/build-with-claude/claude-in-microsoft-foundry |
| OpenRouter | `anthropic/claude-fable-5.2` |  | https://openrouter.ai/anthropic/claude-fable-5.2 |
| Web app (all paid plans), Claude Code |  |  | https://claude.ai |

## Pricing

- Input: $8 per 1M tokens
- Output: $40 per 1M tokens
- Cache write, 5 minutes: $10 per 1M tokens
- Cache write, 1 hour: $16 per 1M tokens
- Cache read: $0.20 per 1M tokens
- Unit as published: per 1M tokens (USD); fast mode (up to 2.5x faster output) $16 input / $80 output
- Source: https://platform.claude.com/docs/en/about-claude/pricing
- Last checked: 2026-11-09

## Specs

- Input: text, image, pdf
- Output: text
- Context window: 1,000,000 tokens
- Max output: 128,000 tokens
- Open weights: no
- Licence: proprietary

## What it has not seen

- Published training cutoff: 2026-06 (June 2026).
- AI events after it on Post-Cutoff: 1,712 (388 major, 79 historic), as of 2026-11-09.
- Already missing at launch: 1,709.
- Briefings of the news since June 2026: https://postcutoff.com/briefings/cutoff-2026-06-s.md (short), https://postcutoff.com/briefings/cutoff-2026-06-m.md (medium), https://postcutoff.com/briefings/cutoff-2026-06.md (full).
- Its card, with five things it has never heard of: https://postcutoff.com/you/claude-fable-5-2/

## What stands out

- Top agentic-coding score at launch: FrontierCode 2.0: 49.1% at Xhigh effort, the highest of any released model Anthropic tested (Opus 5.5 45.5%, GPT-6 Astra 42.8%, Gemini 4 Carbon 41.1%); at High effort it matches Opus 5.5's best for about half the cost per task. Vendor numbers. Source: https://www.anthropic.com/claude-fable-and-mythos-5-2
- Fermat's Last Theorem formalized in 30 hours (first): About 1,000 Fable 5.2 agents fully formalized FLT in Lean in 30 hours with minimal human intervention (the first formalization by an unreleased Claude took 11 days); Opus 5.5 with the same setup reached 26% in 18 hours. Source: https://www.anthropic.com/research/formalizing-fermats-last-theorem
- Robot control without robotics training: RoboDojo Sim (simulated two-arm robot from camera images): 64.9% at Max effort, versus 39.4% for Opus 5.5, 34.2% Gemini 4 Argon and 29.8% GPT-6 Astra in Anthropic's runs; beats most purpose-built robot-control systems on the public board. Source: https://robodojo-benchmark.com/
- Hardware design end to end: From a one-paragraph prompt it specified, sourced parts for, routed four-layer boards and wrote firmware for three devices (voice hub, USB speakerphone, handheld pet) built with Diode; all worked without electrical changes. BenchCAD 2.1: 83.5 vs 65.3 for Opus 5.5. Source: https://www.anthropic.com/news/model-hardware-standard-research-preview
- Open physics problems: Solved two open problems on holographic black holes (Hawking–Page transition at weak coupling is first order; the exact limit of the modular bootstrap slope), posted on Hexagon as 2611.00175 and 2611.00174. Source: https://hexagonmath.org/2611.00175v1
- Fewer safety-classifier blocks: About 55% fewer blocks than Fable 5.1 in early testing (60% fewer in cyber, 40% in biology); blocked cyber queries are answered by Opus 4.8 and biology ones by Opus 5. Source: https://www.anthropic.com/claude-fable-and-mythos-5-2

## Notes

Same underlying model as Claude Mythos 5.2 (Mythos without the cyber/biology classifiers, for verified organizations). Up to 30% faster output than Fable 5.1; optional fast mode up to 2.5x. Default effort: high on the API, medium in Claude apps and Claude Code. Retirement not before 2027-11-09. Anthropic still recommends Opus 5.5 as the starting point for most coding and knowledge work. AI-R&D-specific capabilities (kernel optimization, LLM architecture work) were deliberately held at about Opus 5.5 level during training; transparent fallbacks to Opus 5 on some of those topics. Strongest anti-distillation protections; preserved thinking now applies to all accounts. 30-day retention for safety monitoring unless the customer uses Enterprise Frontier Safeguards (ZDR-compatible).

Anthropic's strongest widely released model from 9 November 2026, replacing Fable 5.1 at a lower price ($8/$40 instead of $10/$50). Best for long-horizon, open-ended technical work and as an Advisor or Orchestrator for smaller Claude models. Handle `stop_reason: "refusal"` from the safety classifiers.

Launch benchmarks (Anthropic, safeguards on, highest effort unless noted): FrontierSWE v2 69.9%, AutomationBench 51.4%, Terminal-Bench-Science 0.1 74.6%, OSWorld 2.1 (offline, partial) 89.3%, WANDR 82.1%, Remote Labor Index 41.3%, Humanity's Last Exam (tools) 74.2%, Chartography 73.4%, AA-Briefcase 66.1% rubric / 1785 Elo. Independent leaderboards had not published results on launch day.

```bash
curl https://api.anthropic.com/v1/messages \
 -H "x-api-key: $ANTHROPIC_API_KEY" -H "anthropic-version: 2023-06-01" -H "content-type: application/json" \
 -d '{"model":"claude-fable-5-2","max_tokens":16000,"messages":[{"role":"user","content":"Hello"}]}'
```

Model page: https://postcutoff.com/m/claude-fable-5-2/