Model components / Fine-tuned / chat models

Claude Fable 5

Anthropic's first Mythos-class model released for general availability in June 2026, sitting above Opus in capability. It shares weights with Claude Mythos 5 but wraps them in safety classifiers that route sensitive cybersecurity, biology, chemistry and distillation queries to Opus. It is built for long-horizon agentic coding, knowledge work, vision and scientific research, with a one-million-token context window and up to 128K output.

Designated a Covered Model, which carries 30-day mandatory data retention. Verified 2026-08-13 via the Fable 5 launch post, the Claude platform documentation and the Claude Fable product page.

Openness

1 high confidence

1.0

weights: closed
data: closed
code: closed
license: Proprietary(API-only)

No weights, data or code released. Served through the Anthropic API only, so fully closed on every dimension. Recorded here rather than among the pretrained base models because a closed model has no observable base checkpoint to score.

https://www.anthropic.com/news/claude-fable-5-mythos-5 recorded 2026-08-13
Mythos-class flagship, proprietary, claude-fable-5 via API and cloud marketplaces only
https://platform.claude.com/docs/en/about-claude/models/introducing-claude-fable-5-and-claude-mythos-5 recorded 2026-08-13
model table - "Claude Fable 5 / claude-fable-5 / Anthropic's most capable widely released model"; "both are designated Covered Models" with 30-day data retention and no zero-data-retention option; access is by API model ID, with no weights, corpus or training code offered

Adoption

4 low confidence

4.0

A directional read on market position rather than a measured usage figure. Anthropic's most capable generally available model, with day-one enterprise testimonials (Stripe, GitHub, Cursor, Cognition) and a staged subscription rollout. Capped at 4 rather than 5 because it launched in June 2026 and sustained reach is still unproven.

https://www.anthropic.com/news/claude-fable-5-mythos-5 recorded 2026-08-13
launch testimonials (Cursor and others) and staged availability; no usage or user figure anywhere on the page
https://openrouter.ai/rankings recorded 2026-08-13
"Live LLM rankings based on benchmarks and real data from millions of people using models through OpenRouter" - the leaderboard is client-rendered, so a plain fetch cannot show which models rank where

Capability

5 high confidence

5.0

Anthropic-reported: highest among frontier models on Cognition FrontierCode (even at medium effort); top on Hebbia Finance Benchmark; SOTA on CursorBench and FrontierBench per partner testing; leads Opus 4.8 on long-horizon coding, knowledge work, vision, and memory tasks. Vendor-reported; underlying Mythos weights with classifier fallback on ~5% of sessions.

https://www.anthropic.com/news/claude-fable-5-mythos-5 recorded 2026-08-13
on Cognition's FrontierCode evaluation "Fable 5 scores highest among frontier models, even at medium effort"; "On Hebbia's Finance Benchmark for senior-level reasoning, Fable 5 has the highest score of any model"; Cursor - "Claude Fable 5 is the state of the art model on CursorBench"
https://www.anthropic.com/claude/fable recorded 2026-08-13
"Claude Fable 5 is our most capable generally available model. It's best for ambitious, long-running, asynchronous work"; long-horizon agentic flagship positioning

Verified 2026-08-13