On June 9, 2026, Anthropic launched Claude Fable 5 (API model ID: claude-fable-5), which it calls its most powerful generally available model to date, and simultaneously reset the 5-hour and weekly rate limits for all users.
June 9, 2026 · Anthropic
Claude Fable 5 goes general — the first Mythos-class model built for long-horizon agentic work
Anthropic's top-tier model arrives across web, API, Bedrock, Copilot and Snowflake — included at no extra cost through June 22 — and resets all 5-hour and weekly rate limits the same day. It's tuned so its lead widens on longer tasks, and Anthropic says it can "run for days."
Senior engineer evaluation — score out of 100
Fable 5 reports SOTA results on nearly all benchmarks, with particular strength in software engineering, knowledge work, scientific research and vision.
95%+
of sessions handled by Fable 5 itself
<5%
high-risk queries routed to the lower-risk Opus 4.8 "fallback"
30 days
data retention for safety classifiers
Built-in safeguard — the fallback guard
High-risk queries — cybersecurity, biology and chemistry, distillation — are switched to the lower-risk model, making Fable 5 the first Mythos-class model hardened for general use.
The praise
Top-class coding; excels at large refactors and long-running projects
Lead widens on longer tasks — "can run for days"
Rate reset lets teams seriously test it on production code
The concerns
Slow and expensive — aimed at advanced agent builders
Heavier consumption on Pro/Max plans
Some performance dips on machine-learning tasks
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article…
