Claude Fable 5

The first publicly released model in Anthropic's Mythos class — a locked-down, broadly-available counterpart to the more capable, restricted Claude Mythos 5 released alongside it. AA Intelligence Index v4.1: 60 — the #1 model on AA's index, ahead of Opus 4.8 (56) and GPT-5.5 (55). State-of-the-art on nearly all tested capability benchmarks (software engineering, knowledge work, vision, scientific research), scoring 10%+ above Opus 4.8 on some.

Public release is gated by domain safeguards: on high-risk queries (e.g. cybersecurity, biology) it blocks and falls back to Opus 4.8 for a safe answer. Priced at $10/M input, $50/M output (same as Mythos 5; less than half of the earlier Mythos Preview, but well above standard Opus). Notably, a US government export directive temporarily forced it offline within days of launch. API name claude-fable-5.