Claude Fable 5.2 is Anthropic's new model, released on 9 November 2026. It is built for demanding reasoning and agent work that runs for hours. At Anthropic it costs 20% less than Fable 5.1, and on Kumo it costs $5.7 per million input tokens and $28.5 per million output tokens, almost 29% below the official price.
Context
1M tokens
up to 128K tokens in one answer
Price on Kumo
$5.7 / $28.5
input / output per 1M tokens
Knowledge
June 2026
reliable cutoff, per Anthropic
Released
9 Nov 2026
on Kumo since 10 November
What Claude Fable 5.2 is
Fable is Anthropic's top generally available line. Version 5.2 is designed for work that does not fit into one short request: agent sessions lasting hours, multistep research, and analysis carried through to a finished document, spreadsheet or deck.
Claude Mythos 5.2 launched alongside it. It is the same model without some of the cybersecurity and biology safeguards, and it is available only to organizations verified by Anthropic. In Fable 5.2, requests from those areas that the safety classifiers flag are automatically handed to less capable models. Anthropic reports about 55% fewer such blocks than Fable 5.1.
Anthropic also reports outputs up to 30% faster than Fable 5.1, and at medium effort the model uses fewer tokens per task. There is also an orchestrator role: Fable 5.2 can lead the work and hand simple steps to smaller models.
What the measurements show
Every figure below was published by Anthropic itself. We found no independent measurements of Fable 5.2 at the time of writing, so treat them as the vendor's claims. The model does not lead everywhere: on AutomationBench, for example, Gemini 4 Carbon scored 53.9% against 51.4% for Fable 5.2.
| Test | Fable 5.2 | Fable 5.1 | Opus 5.5 | GPT-6 Astra |
|---|---|---|---|---|
| FrontierCode 2.0 | 49.1% | 32.4% | 45.5% | 42.8% |
| FrontierSWE v2 | 69.9% | 56.3% | 62.3% | 65.5% |
| Terminal-Bench-Science 0.1 | 74.6% | 52.6% | 58.7% | 68.1% |
| AutomationBench | 51.4% | 31.4% | 42.5% | 41.4% |
| OSWorld 2.1 | 89.3% | 83.5% | 87.2% | 79.5% |
| Remote Labor Index | 41.3% | 17.9% | 21.3% | 20.8% |
| Humanity's Last Exam, with tools | 74.2% | 65.6% | 67.7% | 57.2% |
| WANDR | 82.1% | 68.7% | 72.3% | — |
What Claude Fable 5.2 costs
| Where | Input | Output | Cache read | Cache write |
|---|---|---|---|---|
| Anthropic | $8 | $40 | $0.2 | $10 |
| Kumo, paid from balance | $5.7 | $28.5 | $0.1 ($0.143) | $7.1 ($7.125) |
In Kumo token packages the model is charged with a ×19.38 multiplier. For comparison, Fable 5.1 on Kumo costs $6 / $30 with a ×20.4 multiplier, so the new version is 5% cheaper than the previous one. Re-reading the same context costs far less than ordinary input, so long cached sessions come out noticeably cheaper.
Claude Fable 5.2 in the Kumo catalogue
The model card with its current price is below; it leads straight to setup.
What it is good for
- Large-scale development: loosely defined tasks across several repositories, with the model checking its own results.
- Long agent sessions where the model coordinates subtasks and carries the work through.
- Multistep research and financial analysis ending in a finished report.
- Data work and machine learning: measure, try, improve loops.
- Where price matters more than peak quality, Opus 5.5 or Sonnet is the better choice: Fable is meant for the hardest tasks.
What Kumo has verified
Before launch Kumo runs probe requests against every model. For Claude Fable 5.2, plain answers, streaming and tool calling are confirmed over the Anthropic Messages and Chat Completions protocols, and reasoning control over Anthropic Messages.
How to connect Claude Fable 5.2
- Sign up in the Kumo console and top up your balance.
- Create an API key and set a spending limit on it.
- Point any OpenAI-compatible client or SDK at the gateway https://api.kumorouter.com/v1.
- Set the model to claude-fable-5-2 and send your first request.
curl -s https://api.kumorouter.com/v1/chat/completions \
-H "Authorization: Bearer $KUMO_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "claude-fable-5-2",
"messages": [
{"role": "user", "content": "Find the bug in this function and suggest a fix: ..."}
],
"max_tokens": 4096
}'
Notes for developers
- Do not send temperature, top_p or top_k: per Anthropic's docs some values return a 400 error.
- Forced tool use (tool_choice types any and tool) returns a 400 error on this model.
- Adaptive thinking is always on, and the effort parameter sets its depth. Anthropic's default is high.
FAQ
- How does Fable 5.2 differ from Fable 5.1?
- According to Anthropic it is stronger on coding and agent tests, answers up to 30% faster and blocks about 55% fewer requests. The official price fell from $10 / $50 to $8 / $40 per million tokens.
- How does Fable 5.2 differ from Opus 5.5?
- Fable 5.2 is stronger on the hardest tasks in Anthropic's tests but noticeably more expensive: on Kumo $5.7 / $28.5 against $2.2 / $11 for Opus 5.5. For everyday work Opus 5.5 usually pays off better.
- Why is Kumo cheaper than Anthropic?
- That is Kumo's price for Claude Fable 5.2: $5.7 against $8 per million input tokens and $28.5 against $40 for output. You pay from your Kumo balance, with no separate Anthropic account.
- Can I use the model in my own agent?
- Yes. Kumo accepts requests over the Chat Completions and Anthropic Messages protocols, so a client that speaks either one works. Set the gateway address, your Kumo key and the model claude-fable-5-2.
Sources
- Announcement and measurements: Introducing Claude Fable 5.2 and Claude Mythos 5.2, Anthropic, 9 November 2026.
- Vendor specs and prices: Claude Fable 5.2 documentation.
- Kumo prices and specs: models page, price list v48.