AI models · Anthropic

Claude Fable 5.2

Anthropic released Claude Fable 5.2 on 9 November 2026 as the successor to Fable 5.1. It scores 65 on the independent Artificial Analysis index, in first place ahead of Gemini 4 Carbon and Opus 5.5. It costs $8 input and $40 output per million tokens, a fifth less than Fable 5.1. For everyday work, though, Anthropic still recommends starting with Opus 5.5, which costs half as much, and keeping Fable for tasks Opus cannot handle.

Strengths

  • It scores 65 on the Artificial Analysis Intelligence Index v5.0.1 at max effort, the highest of any model. Gemini 4 Carbon scores 60 and Opus 5.5 59.
  • It is cheaper than Fable 5.1: $8 instead of $10 for input and $40 instead of $50 for output, and cached input costs $0.20. Anthropic says it also writes up to 30% faster.
  • It leads Anthropic's coding tests: 49.1% on FrontierCode 2.0, against 45.5% for Opus 5.5 and 42.8% for GPT-6 Astra. At high effort it matches Opus 5.5 at max for roughly half the cost per task.
  • A big jump in autonomous office and research work: 41.3% on the Remote Labor Index against 21.3% for Opus 5.5, and 74.2% on Humanity's Last Exam with tools against 67.7% (Anthropic's figures).
  • Anthropic says safety filters step in needlessly about 55% less often than on Fable 5.1. Fast mode writes up to 2.5 times faster at double the price.

Weaknesses

  • It still costs twice as much as Opus 5.5 ($4 / $20), which Anthropic recommends as the main model for everyday work.
  • It is very wordy: it used 275 million output tokens on the Artificial Analysis tests, while a typical model uses about 73 million. The real cost per task goes up.
  • On Pro and Team Standard it is outside the plan limits and runs on credits only, except for a launch week until 16 November 2026. It is missing from the free plan. On Max, Ultra and Team Premium it can use at most half of the weekly limit.
  • Safety filters reroute penetration testing, exploits and binary vulnerability work to Opus 4.8 and high-risk biology queries to Opus 5. Forcing tool use through the API returns an error.
  • It requires 30-day data retention, so companies on zero data retention (ZDR) cannot use it without Anthropic's approval. Anthropic has not published Terminal-Bench 4.0 or SWE-bench results.

Best for

  • The hardest coding tasks that run for hours
  • Deep multi-step research
  • Long-running autonomous AI agents that direct other agents
  • Complex work with documents, spreadsheets and slides
  • Tasks where Opus 5.5 falls short

Not ideal for

  • Everyday chat
  • High-volume work where cost matters
  • Apps that need fast replies
  • Offensive security testing, safety filters reroute it
  • Free plan users

How to use it in Europe

In Europe you get it in the Claude app on the web, desktop and mobile with a Max or Ultra plan or a premium Team or Enterprise seat, using at most half of your weekly limit. On Pro and Team Standard you can only use it with paid credits, and not at all on the free plan. Claude Code needs version 2.1.321 or later. Developers call it through the Anthropic API, Amazon Bedrock, Google Cloud and Microsoft Foundry. The direct API stores data in the US, with no EU region to choose. Anthropic says text from models released on or after 2 August 2026 carries an invisible watermark because of the EU AI Act.

What is coming

Confirmed by vendor According to Anthropic, this is the same model as Claude Fable 5.2, reserved for vetted organizations in cybersecurity and biomedical research. Anthropic has not announced access for everyone. When: no date yet, Anthropic is gradually expanding its verification programs →

Price compared with similar models

0.3 1 3 10 30 Ministral 3 14B0.2 Llama 4 Scout0.24 Qwen3.8-Flash0.47 Claude Haiku 5.50.5 Muse Ember0.5 Nemotron 3.5 Super VL0.5 GPT-6 Luna0.5 GLM-5.3-Flash0.5 Mistral Small 40.6 Llama 4 Maverick0.65 Codestral 25.080.9 Beam 1.01 DeepSeek V4.1 Flash1.2 Muse Glimmer 30B1.2 Qwen3.7-Plus1.6 Grok Build 0.12 Mistral Large 42.1 Gemini 3.5 Flash-Lite2.5 Grok 4.32.5 Qwen3.8-27B3 Gemini 3.8 Flash3.75 DeepSeek V4 Pro3.96 Muse Spark 1.34.25 Muse Spark Open4.25 GLM-5.34.4 GPT-6.1 Terra5 Qwen3.8-Max6 Qwen3.8-2.4T-A95B6 Grok 4.76 Claude Sonnet 5.510 Gemini 4 Argon10 Kimi K3.1 Pro10 GPT-6.1 Sol10 Kimi K315 Claude Opus 5.520 Muse Wonder20 Claude Fable 5.240 GPT-6 Astra50
Output price in USD per 1M tokens (log scale). Lighter bars are open-weight models.

Benchmark results

BenchmarkScoreSource
Artificial Analysis Intelligence Index v5.0.1 (max effort)65 (#1 of 52)artificialanalysis.ai ↗, November 2026
FrontierCode 2.0, xhigh effort49.1%anthropic.com ↗ (vendor-reported), November 2026
FrontierSWE v269.9%anthropic.com ↗ (vendor-reported), November 2026
Humanity's Last Exam, with tools74.2%anthropic.com ↗ (vendor-reported), November 2026
Remote Labor Index, high effort41.3%anthropic.com ↗ (vendor-reported), November 2026

Benchmarks cover only a narrow slice of ability and vendors often report them at their best settings. Treat them as a rough guide.

More models from this vendor

  • Claude Opus 5.5: Anthropic's main model for demanding work, coding and long documents, and the one Anthropic recommends starting with.
  • Claude Sonnet 5.5: Claude's balanced model for everyday work and coding. It is faster than Opus 5.5 and costs half as much.
  • Claude Haiku 5.5: Claude's cheapest and fastest model, built for classification, data extraction and large volumes of simple tasks.

Similar models from other vendors

  • gpt-oss-120b (OpenAI): OpenAI's open model that a company can run on its own server when data must not leave its infrastructure.

Sources

Prices and models change often. The data comes from official vendor pricing pages and documentation, and every page shows when it was last checked. Before a larger purchase, confirm the price with the vendor.

← All AI models

Not sure which model fits your business?

I can help you pick a model and plan based on what you actually need, set it up safely for company data and train your team to use it well. Let’s start with an intro consultation.

See pricing [email protected]