Claude Fable 5.1

Overview. Claude Fable 5.1 is Anthropic's proprietary, closed-weight frontier reasoning model, released as an incremental successor to Claude Fable 5. It pairs a 1,000,000-token context window with mandatory adaptive reasoning that scales from low to max effort on every request, and accepts text and image input while producing text-only output. Developers reach for it when a task needs deep multi-step reasoning over very large inputs rather than a fast, low-effort response.

Key strengths. Artificial Analysis places Claude Fable 5.1 in the frontier quality tier, with an Intelligence Index of 65.7 at max effort and a Coding Index of 81.6, indicating particular strength on code-generation and code-reasoning tasks. Its Agentic Index of 61.3 supports multi-step tool-use and planning workloads. The model's always-on reasoning process (mandatory, not optional) is the defining trait behind these scores.

When to use it. Use Claude Fable 5.1 for complex coding tasks, multi-step agentic workflows, and analysis over very large documents or codebases that benefit from the 1,000,000-token context window. Its native image input also makes it suitable for tasks that combine visual and text reasoning, such as interpreting diagrams or screenshots alongside written instructions. Because reasoning effort is adjustable from low to max, it can be tuned for either deeper analysis or faster turnaround within the same model.

Limitations. Reasoning is mandatory and cannot be disabled, which adds latency compared to non-reasoning chat models, even though the model remains usable in interactive settings. It does not support a `temperature` parameter, and output is text-only — it cannot generate images or audio. Its training data cutoff is June 2026.

Sovereignty note. Claude Fable 5.1 is built by Anthropic and served under Anthropic's Acceptable Use Policy; no distinct licence exists for this model. The available research does not confirm endpoint data residency or CLOUD Act exposure, so no Sovereign designation applies.