Claude Fable 5

Claude Fable 5 is the first model in the Claude 5 generation to reach general availability, and it is the one Anthropic points to for creative and long form work. Where Claude Opus 4.8 is tuned for rigorous reasoning, Fable 5 is tuned for voice, structure and the kind of sustained output that longer projects demand.

The difference shows up most in writing. Fable 5 holds a consistent tone across thousands of words, keeps track of characters, arguments and threads that earlier models tended to drop, and revises its own drafts when you ask it to rather than starting over. That makes it a strong fit for editorial work, documentation, scripts, marketing copy and any task where the reader will notice whether the writing actually reads well.

It is also a capable agent. Fable 5 plans multi step tasks, calls tools, and works through a task list without needing a nudge at every turn, which is why teams building assistants and workflow automation have adopted it quickly. A 500,000 token context lets it take in long briefs, full style guides or an entire manuscript at once, and extended thinking is available when a task benefits from planning before the model commits to an answer.

On raw reasoning benchmarks Fable 5 sits just below Opus 4.8, which remains the family leader for graduate level science and the hardest coding problems. That gap is deliberate. Fable 5 trades a little benchmark performance for a lower price and a more natural writing style, so it often gives the better result on open ended work while costing a fraction of the frontier model. For high volume or latency sensitive jobs, Claude Haiku 4.5 stays the efficient option.

Fable 5 shares the same safety training and guardrails as the rest of the lineup. It is separate from the gated Claude Mythos model, which Anthropic restricts to vetted security and research use. Fable 5, by contrast, is open to any developer through the API and the Claude apps.

If your work is mostly writing, editing, long documents or agent building, Fable 5 is usually the model to start with, and you can move individual hard requests up to Opus 4.8 when you need the extra reasoning. Compare the full family on the Claude models page, or read our reporting on the Claude 5 launch for background on how the generation arrived.

Vision rounds out the picture. Fable 5 reads images, screenshots, charts and document scans, so it can describe a design, pull the structure out of a slide, or turn a whiteboard photo into clean notes. Combined with the long context, that makes it a practical tool for turning messy source material into finished writing in a single pass, rather than copying text back and forth between separate tools.

Fable 5 supports vision, prompt caching and tool use, and is available in the Anthropic API, the Claude apps and the main cloud platforms, so you can prototype in the assistant and ship the same model in production.