Initial impressions of Claude Fable 5
Simon Willison · Simon Willison · 2026-06-09
Simon Willison shares first-day impressions of Anthropic's Claude Fable 5, describing it as a large, expensive frontier model with strict safety guardrails, a 1 million token context window at $10/$50 per million tokens, and strong agentic coding capability demonstrated by generating a Python WASM sandbox and a major LLM library release in a single session.
Appears in
Extraction
Topics: claude-fable-5frontier-modelsllm-pricingai-safety-guardrailsagentic-coding
Claims
- Claude Fable 5 offers the same capabilities as Claude Mythos 5 but with stricter safety classifiers that trigger frequently enough to warrant new API fallback mechanisms.
- Claude Fable 5 has a 1 million token context window, 128,000 maximum output tokens, a January 2026 knowledge cutoff, and is priced at $10/million input and $50/million output tokens — twice the price of Opus 4.x.
- Claude Fable 5 demonstrates substantially broader factual knowledge of open-source projects than Opus 4.8 when queried without search tools.
- Willison used Fable 5 to nearly fully implement LLM 0.32a3, including new API design, tests, documentation, and five library improvements, in a single day of work.
- Willison spent $110.42 worth of Fable 5 tokens in a single day on a $100/month subscription.
Key quotes
The best way to describe Fable is that it feels big. Not just in terms of speed and cost, but also in how much it knows.
I'm really impressed with the quality of API design, tests, code and documentation that Fable put together for this. I spent several hours on it today, but it feels like several days' worth of work.
As is frequently the case with current frontier models the challenge is finding tasks that it can't do.