Claude Fable 5

• 🆕 First publicly available model from Anthropic's Mythos class
• 📏 1M-token context with up to 128K output tokens
• 🧠 Always-on adaptive thinking for long-horizon agentic work
• 👁️ Multimodal vision capabilities for reasoning over images
• 🔧 Coding, function calling, and web search built in
• 🏢 Available via the Claude API and cloud platforms
• 📚 Anthropic reports gains over Opus 4.8 on agentic coding
• 🔒 Ships with additional Mythos-class safety safeguards

Anthropic

Anthropic PBC is an American artificial intelligence company headquartered in San Francisco. Structured as a public benefit corporation, the lab develops large language models under the Claude name, with a research emphasis on building reliable, steerable,…

Claude Fable 5 is Anthropic's most capable widely released model and the first commercially available member of its top-tier Mythos class, released in 2026.[2][3] Positioned above the Opus tier, it pairs a 1M-token context window with up to 128K output tokens, always-on adaptive thinking, and multimodal vision, and is designed for demanding reasoning and long-horizon agentic work.[2][3]

Compared with its same-family predecessor Claude Opus 4.8 — Anthropic's previous flagship, released just weeks earlier — Fable 5 sits a tier higher and is presented as an advance on agentic coding, reasoning, and visual understanding, with always-on adaptive thinking enabling longer, asynchronous task execution.[1][2] Anthropic reports gains over Opus 4.8 on agentic coding in its announcement, though specific figures should be verified against the company's own model card.[2]

As a Mythos-class model, Fable 5 ships with additional safety safeguards, and its unrestricted counterpart, Mythos 5, is offered through more limited access channels rather than general availability.[2][3] Within Anthropic's broader lineup it succeeds the Opus generation, including Claude Opus 4.7 and Claude Opus 4.6, extending the family's emphasis on coding, tool use, and extended autonomous workflows.[3]