Anthropic Releases Claude Fable 5 with Extended Context Window
The latest model from Anthropic pushes context limits further while improving instruction-following and reducing hallucinations in long-form tasks.
Anthropic has released Claude Fable 5, its most capable model to date, featuring a dramatically expanded context window and significant improvements to instruction-following accuracy. The release marks another step forward in the company's effort to build AI systems that are both more useful and more reliable in demanding real-world applications.
What's new in Fable 5
The headline feature is an extended context window that allows the model to process substantially longer inputs in a single pass — useful for tasks like summarizing lengthy legal documents, analyzing full codebases, or maintaining coherence across book-length conversations. Anthropic says Fable 5 also shows measurable gains on multi-step instruction following, meaning it's less likely to drop constraints halfway through a complex task.
Reduced hallucinations in long-form tasks
One of the persistent challenges with large language models is maintaining factual accuracy across long outputs. Fable 5 reportedly improves on this front, particularly in tasks that require synthesizing information from multiple parts of a long context. Internal benchmarks show a notable reduction in confident-but-wrong responses when the model is asked to reason across documents spanning tens of thousands of tokens.
Availability and pricing
Claude Fable 5 is available immediately via the Anthropic API and through Claude.ai. Pricing follows a tiered structure based on input and output tokens. Enterprise customers with existing contracts will need to update their API calls to specify the new model ID. Anthropic has also confirmed that Fable 5 will be accessible through Amazon Bedrock and Google Cloud Vertex AI in the coming weeks.
What this means for developers
For teams already building on Claude, the upgrade path is straightforward — swap the model identifier and evaluate outputs on your existing test cases. The expanded context window opens new use cases that previously required chunking strategies or retrieval-augmented generation workarounds. Developers working on document-heavy applications will likely see the most immediate benefit.