Anthropic released Claude Fable 5.1 this week as its latest high-end model for complex tasks that require sustained reasoning, according to Tom's Guide. The model is built around agentic work (planning and using tools with less supervision), with a 1-million-token context window for large codebases, long documents, and research-heavy projects. Anthropic says Fable 5.1 improves coding and long-running tasks while lowering some agentic workload costs through cheaper caching.

Tom's Guide tested Fable 5.1 with a project-style prompt rather than a clean request. The suggested prompt asks users to describe a real project, paste their current code, outline, or plan, and ask Claude to identify flawed thinking, missing pieces, and then refactor or redesign the work while explaining each decision. The test is meant to show whether Fable 5.1 can infer the user's goal, evaluate existing assumptions, and decide which choices should stay or change.

The article says the prompt contains several hidden tasks: understanding the project, finding weaknesses the user missed, judging trade-offs across the work, and tracking redesign decisions. The line asking Claude to explain its reasoning is presented as central because it lets users see not only what the model recommends but why it chose one path over another.

Tom's Guide says Fable 5.1 can be tested on more than software. Users could give it code, a business plan, a novel outline, a research project, a presentation, or a home renovation plan, especially when earlier decisions may no longer make sense. Amanda Caswell wrote that Fable 5.1 gave thorough feedback and line edits on a manuscript, questioned assumptions, suggested structural changes for clarity, and found a chapter problem she had not noticed. She asked it to list changes instead of rewriting the manuscript directly, and said its explanations mattered as much as the edits.

Anthropic released Claude Fable 5.1 this week as its latest high-end model for complex tasks that require sustained reasoning, according to Tom's Guide. The model is built around agentic work (planning and using tools with less supervision), with a 1-million-token context window for large codebases, long documents, and research-heavy projects. Anthropic says Fable 5.1 improves coding and long-running tasks while lowering some agentic workload costs through cheaper caching.

Tom's Guide tested Fable 5.1 with a project-style prompt rather than a clean request. The suggested prompt asks users to describe a real project, paste their current code, outline, or plan, and ask Claude to identify flawed thinking, missing pieces, and then refactor or redesign the work while explaining each decision. The test is meant to show whether Fable 5.1 can infer the user's goal, evaluate existing assumptions, and decide which choices should stay or change.

The article says the prompt contains several hidden tasks: understanding the project, finding weaknesses the user missed, judging trade-offs across the work, and tracking redesign decisions. The line asking Claude to explain its reasoning is presented as central because it lets users see not only what the model recommends but why it chose one path over another.

Tom's Guide says Fable 5.1 can be tested on more than software. Users could give it code, a business plan, a novel outline, a research project, a presentation, or a home renovation plan, especially when earlier decisions may no longer make sense. Amanda Caswell wrote that Fable 5.1 gave thorough feedback and line edits on a manuscript, questioned assumptions, suggested structural changes for clarity, and found a chapter problem she had not noticed. She asked it to list changes instead of rewriting the manuscript directly, and said its explanations mattered as much as the edits.