Claude Fable 5.1
Claude Fable 5.1 is a closed AI model from Anthropic with an index score of 189 (GPT-4 = 100), rank 11 of 217.
189173–205Index score#11 of 217 models
$10/ $50Price per million tokensinput / output · via OpenRouter
1MContext windowtokens it reads at once
1 Sep 2026ReleasedReplaced by Claude Fable 5.2
Where it sitson the frontier
Among the first models to reach this score.
Test resultsby area · tick = best by any model
Reasoning & knowledgeevidence
Mathsevidence
Codingevidence
Agentsevidence
Professional workevidence
Novel problemsevidence
Long tasksno evidence yet
Third-party results only, at the model’s best standard setting. Hover a bar for who ran it.
What the lab says3 of 8 also run by third parties
| Test | Anthropic says | Third parties |
|---|---|---|
| Terminal-Bench Science 0.1Claude Code harness, 3 tries per task. Anthropic gives a margin of ±3.5–4.5 points. | 52.6% | 40.0%Vals AI |
| Terminal-Bench 4.0 | 55.8% | 58.1%Vals AI |
| Humanity's Last Examno tools | 60.9% | 46.5%Scale AI |
| GDPval-AA v2 | 1853 Elo | not run yet |
| OSWorld 2.0partial credit | 77.9% | not run yet |
| Humanity's Last Examwith tools | 65.0% | not run yet |
| AutomationBench | 31.4% | not run yet |
| CursorBench 3.2.0 | 73.4% | not run yet |
From Anthropic’s launch post, Claude Fable 5.1 and Mythos 5.1, copied from the table in the post. Shown, never scored; settings can differ from the third-party run.
What others sayrank on 13 leaderboards
The Model Indexthird-party tests, our method#11of 217 · 189
LMArena · Mathsmaths questions in chat#2of 346 · 1526LMArena · Textoverall chat quality, judged by people#7of 362 · 1502LMArena · Instruction followingfollowing detailed instructions#8of 362 · 1501LMArena · Creative writingcreative writing#8of 360 · 1490LMArena · Documentsworking with long documents#1of 40 · 1515LMArena · Long promptslong, complex requests#9of 340 · 1509Epoch Capabilities Indexoverall capability from many benchmarks#8of 231 · 167.1LMArena · Codingcoding questions in chat#15of 357 · 1533LMArena · WebDevbuilding web apps and front ends#7of 133 · 1743LMArena · Visionunderstanding images#13of 145 · 1289LiveBench · Overalloverall, on regularly refreshed questions#4of 38 · 84LMArena · Searchanswering with web search#6of 41 · 1237Best attop three among current models
Where to use itprices checked 18 Nov 2026
API providers 4 · per 1M tokens, in / out
Host prices from OpenRouter; going direct can differ.
Facts and sources
Released1 Sep 2026
Announcementanthropic.com ↗
Release entryClaude Fable 5.1 + Mythos 5.1
Scored asClaude Fable 5.1
Output speed–
Hugging Faceclosed weights