Claude Fable 5.1
Senior Research Engineer at Anthropic | Long-horizon agentic work and hard reasoning | Cheaper cache reads than Fable 5
The basics
Checked Oct 22- Status
- Model ID
- Price
- Context
- Supported to
Updates
Strengths and weaknesses
Strengths
Weaknesses
- !
- !
- !
- !
Role
Long-horizon reasoning model at Anthropic
Sep 2026 to now · 2 months- Sep 1, 2026: Current Fable model
Jun 2026 to Sep 2026 · 3 months
Works with
From recorded changes
API · 7 shared changes
Agent · 7 shared changes
App · 6 shared changes
Agent · 6 shared changes
Agent · 1 shared change
Similar models
$4 in, $20 out
$10 in, $50 out
$10 in, $50 out
Cost vs peers
Tied for most expensive of 5 similar models by output price.
Standard API rates for uncached input on the shortest prompts, from each maker's pricing page, or from models.dev for models not researched here yet.
Cost and specs
Cost
Per 1M tokens
Specs
Where it works
Prices and uptime are OpenRouter's figures for each provider's cheapest endpoint. Uptime covers the last 24 hours.
Benchmarks
Higher is better. Green is Claude Fable 5.1. Dark marks are similar models.
Independent
Self-reported by Anthropic
Terminal-Bench-Science 0.1
52.6%
Terminal-Bench 4.0
55.8%
CursorBench 3.2.0
73.4%
Chemical and biological capability
CB-1, below the CB-2 threshold, per Anthropic
Alignment risk of catastrophic harm
Low, raised from very low
Critical severity jailbreak found
None, per Anthropic
What people say
Reviews (1)
Lior GrossmanShipping software since 1993, now building with AIFable 5.1 is the most big-picture model I use. I bring it in for broad thinking across my whole business, such as monthly and quarterly strategy reviews. It's fantastic at connecting the dots across different domains and thinking outside the box, and it's good with product decisions and design. In September I also tried it as the orchestrator in Claude Code, with Opus 5.5 writing most of the code, and it was really great as an orchestrator. What did not work: it makes up facts more often than Opus, so anything I'm about to act on from a Fable review gets checked by Opus or Astra first. It's also expensive, so I use it rarely. For page copy, it never beat Opus 5 by more than normal variation, at twice the price.
Oct 31, 2026
Incidents
Incidents per week, last 13 weeks. Most in one week: 4.
- Oct 25Anthropic reported elevated errors across five Claude models30 min
- Oct 25Anthropic reported elevated errors for Claude Fable 5.121 min
- Oct 24Anthropic reported elevated errors for Claude Fable 5 and Fable 5.135 min
- Oct 23Anthropic reported elevated errors for Claude Fable 5.1 and other models31 min
- Oct 5Anthropic reported elevated errors for Claude Mythos 5.1 and Fable 5.130 min
Writing habits
Creative writing · EQ-Bench, among 51 models tracked here
Slop score
8.2
Overused AI words and phrases. Less slop than 98% of tracked models.
Repetition
3.6
Repeated words and ideas. Less repetition than 58% of tracked models.
Story length
5,841
Average characters per story. Longer than 16% of tracked models.