Initial impressions of Claude Fable 5: a capable but slow frontier model
After ~5.5 hours of testing Claude Fable 5, the article describes it as a powerful but slow and expensive frontier model with strict safety guardrails, a 1 million token context window, and pricing at $10/million input and $50/million output tokens — twice the cost of Claude Opus 4.5–4.8.
Score breakdown
Claude Fable 5 represents a new pricing and capability tier in Anthropic's model lineup, introducing both a safety-gated variant and an unconstrained counterpart (Mythos 5) at twice the cost of the Opus 4.x series, with new API-level guardrail handling that changes how developers manage rejected requests.
- 01Claude Fable 5 was tested for ~5.5 hours without early access, described as slow, expensive, and highly capable.
- 02Anthropic claims Fable 5 matches Claude Mythos 5's performance but adds much stricter safety guardrails.
- 03The Claude API gained new mechanisms to notify users when guardrails trigger, plus an option to auto-fall back to another model.
The article covers initial impressions of Claude Fable 5 following approximately 5.5 hours of testing without early access. The model is characterized as "big" — not just in terms of speed and cost, but in the breadth and depth of its knowledge. Anthropic claims Claude Fable 5 delivers the same performance as Claude Mythos 5, but with significantly stricter safety guardrails. Those guardrails activate frequently enough that the Claude API now includes new mechanisms to notify users when they are triggered, as well as a new option to automatically fall back to another model when a request is rejected. Claude Mythos 5, also released the same day, is described by Anthropic as sharing Fable 5's capabilities but without the safety classifiers.
They are priced at $10/million input tokens and $50/million output tokens — exactly twice the price of Claude Opus 4.5/4.6/4.7/4.8 — with no price increase for longer context usage.
Both models feature a 1 million token context window, 128,000 maximum output tokens, and a knowledge cutoff of January 2026. They are priced at $10/million input tokens and $50/million output tokens — exactly twice the price of Claude Opus 4.5/4.6/4.7/4.8 — with no price increase for longer context usage. The article notes the upgrade guide is substantially thinner than the comparable guide for Opus 4.8.
To illustrate Fable 5's knowledge depth, the article presents a side-by-side comparison using the same prompt asking both models to list Simon Willison's open-source projects. Opus 4.8 responded cautiously, naming only a handful of well-known projects with hedged confidence. Fable 5, running without search access, produced a significantly more detailed and chronologically organized list — including projects such as `files-to-prompt`, `datasette-extract`, `LLM`, `symbex`, `ttok`, `strip-tags`, `datasette-lite`, `shot-scraper`, `s3-credentials`, `django-sql-dashboard`, and the Dogsheep suite — with approximate release dates for each.
Key facts
- 01Claude Fable 5 was tested for ~5.5 hours without early access, described as slow, expensive, and highly capable.
- 02Anthropic claims Fable 5 matches Claude Mythos 5's performance but adds much stricter safety guardrails.
- 03The Claude API gained new mechanisms to notify users when guardrails trigger, plus an option to auto-fall back to another model.
- 04Claude Mythos 5, also released the same day, shares Fable 5's capabilities but without the safety classifiers.
- 05Both models have a 1 million token context window, 128,000 maximum output tokens, and a January 2026 knowledge cutoff.
- 06Pricing is $10/million input tokens and $50/million output tokens — twice the price of Claude Opus 4.5/4.6/4.7/4.8.
- 07A side-by-side prompt comparison showed Fable 5 producing a far more detailed and specific list of open-source projects than Opus 4.8.
Topics
Summary and scoring are generated automatically from the original article. We always link back to the publisher and never republish images or paywalled content. Last processed Jun 10, 2026 · 15:34 UTC. How this works →