Anthropic rolls out Claude Fable 5 and Claude Sonnet 5 AI models

Anthropic announced the launch of two new AI models. Claude Fable 5 became available worldwide on 1 July, initially limited to 50 % of weekly usage until 7 July before a credit‑based pricing system takes effect. The model adds a safety classifier to block inappropriate queries, though the company warns of occasional false classifications. Access to the older Mythos 5 model was restored for certain U.S. organisations after government approval, with U.S. Commerce Secretary Howard Lutnick highlighting the partnership’s role in strengthening American AI leadership.

Claude Sonnet 5, released concurrently, is designed for greater autonomy and lower operating costs. It can handle planning, web‑browser and terminal interactions, programming and other complex tasks, achieving performance close to the flagship Claude Opus 4.8. Compared with Claude Sonnet 4.6, it exhibits fewer undesirable behaviours and stronger resistance to prompt‑injection attacks, though it is less capable in cybersecurity‑specific tasks. The model is priced at a promotional $2 per million input tokens and $10 per million output tokens through 31 August 2026, after which rates rise to $3 and $15 respectively.