# Claude Fable 5.2

> Anthropic's flagship model, released November 9, 2026: $8/$40 per million input/output tokens, 1M context, 128K max output, first place on the Artificial Analysis Intelligence Index v5. Cost calculator and access checker on the page. Independent explainer, not affiliated with Anthropic. Updated 2026-11-11. Full page: https://lab.awaara.cn/claude-fable-5-2/

## What it is

- Claude Fable 5.2 is Anthropic's most capable model, built for long-horizon agentic work, hard coding, science, and robotics. Claude Mythos 5.2 is the same underlying model with the cybersecurity and biology safeguards removed, available only to organizations approved through Anthropic's Cyber or Life Sciences Verification Programs.
- Released November 9, 2026. API model ID: claude-fable-5-2. Retirement not sooner than November 9, 2027.
- Specs: 1M-token context window, 128K-token maximum output, June 2026 knowledge cutoff, adaptive thinking always on, default effort high (one step above Opus 5.5 at medium).

## Pricing (Anthropic list prices, November 9, 2026)

- Standard: $8 per million input tokens, $40 per million output tokens. 20% below Fable 5.1 ($10/$50) on every component, including cache and batch.
- Cache: reads $0.20, five-minute writes $10, one-hour writes $16 per million tokens. Batch API: half the standard input and output rates.
- Fast mode: $16 input and $80 output per million tokens, up to 2.5x token generation. Available on the Claude Platform, in Claude Code, on Google Cloud, and on Microsoft Foundry. Batch is not available with fast mode.
- Subscriptions: Fable 5.2 is not on the Free plan. On Pro and standard Team seats it normally runs on usage credits; a launch offer through November 16, 2026 includes Fable models for up to 50% of weekly usage limits, plus one limit reset. Max includes it up to 50% of weekly limits.
- The page's cost calculator prices a monthly workload (input tokens, output tokens, cache percent, fast mode, batch) and compares the same workload on Fable 5.1 and Opus 5.5. The access checker answers Fable vs Mythos for seven situations: API customer, Pro, Team, Max, Enterprise, verified organization, Free.

## Benchmarks (Anthropic-reported, highest effort unless noted)

- FrontierCode 2.0: Fable 5.2 49.1% (Xhigh), Fable 5.1 32.4%, Opus 5.5 45.5%, GPT-6 Astra 42.8%, Gemini 4 Carbon 41.1% (medium effort).
- FrontierSWE v2: 69.9% vs 56.3% (5.1), 62.3% (Opus 5.5), 65.5% (Astra), 61.2% (Carbon).
- AutomationBench: 51.4% vs 53.9% for Gemini 4 Carbon, which leads this one.
- Terminal-Bench-Science: 74.6% (Mythos 5.2: 75.3%). WANDR relaxed: 82.1%. OSWorld 2.1: 89.3% (partial offline set). AA-Briefcase rubric: 66.1% (Elo 1785, behind Opus 5.5 at 1800). Remote Labor Index: 41.3% (high effort). Humanity's Last Exam with tools: 74.2%. Chartography without tools: 73.4%.
- Independent check: Artificial Analysis Intelligence Index v5 ranks Fable 5.2 first at 64.6, ahead of Gemini 4 Carbon at 60.0 and Opus 5.5 at 59. At the highest AA tier a Fable 5.2 task averages $26.50, about 2.6 times Opus 5.5.

## When to use it

- Anthropic's routing: start with Opus 5.5 for most coding and knowledge work, move up to Fable 5.2 for long-horizon agentic work, hard coding, or science, or when Opus at a higher effort still falls short. Fable can also act as advisor (other Claude models consult it) or orchestrator (it delegates simpler tasks to smaller models).
- Fable 5.2 costs twice Opus 5.5 per token. On FrontierCode 2.0 at high effort it scored about level with Opus 5.5 at max effort for roughly half the cost per task.
- Fable's safeguards block most cybersecurity and biology tasks. That work needs Mythos 5.2 and a verified organization.

## Experiments Anthropic cites (prerelease use, untested independently)

- A prototype operating system: over 1,000 subagents, eight days, boots on a laptop.
- Fermat's Last Theorem formalized in Lean in 30 hours (the earlier research version took 11 days).
- Two open problems in black-hole physics solved; three working electronic devices designed from one-paragraph prompts.

## What is still unknown

- Parameter count not published. No open weights, no public license. Real session cost depends on cache hits and effort settings; thinking tokens bill as output.

## Sources

- Anthropic, Introducing Claude Fable 5.2 and Claude Mythos 5.2 (November 9, 2026): https://www.anthropic.com/news/claude-fable-and-mythos-5-2
- Reuters, Anthropic unveils new Fable, Mythos models as competition heats up ahead of IPO (November 9, 2026): https://www.reuters.com/technology/anthropic-unveils-new-fable-mythos-models-competition-heats-up-ahead-ipo-2026-11-09/
- SandBase, Claude Fable 5.2 Is Out: $8/$40, Fast Mode, When to Switch (November 10, 2026): https://blog.sandbase.ai/claude-fable-5-2-release-whats-new-2026/
- SW Architect, November AI Part 1: Claude Fable 5.2 Tops 7 New Models (November 2026): https://www.swarchitect.net/en/ai-launches-2026-11/