"The extended context window enables users to submit far larger datasets within a single request. For developers, this means the ability to load entire codebases-tens of thousands of files, including tests and documentation-while maintaining cross-file awareness. For researchers, it enables synthesis across dozens of lengthy documents, such as academic papers or legal contracts, without losing references along the way."
"The technical implications are significant: larger windows reduce the need to repeatedly retrieve or re-embed content, making multi-day tasks more feasible. However, they also increase the computational load and can make responses less focused if not managed carefully. Some developers argue that scaling context alone is not enough. On Hacker News, user aliljet noted: Allowing me to flood the context window with my code base is great, but short of evals that look into how effective Sonnet stays on track, it is not clear if the value actually exists here."
Claude Sonnet 4 now supports context windows up to 1,000,000 tokens, a fivefold increase over the prior limit. The capability is available in public beta via the Anthropic API and Amazon Bedrock, with Vertex AI support expected soon. The extended context permits single-request submission of much larger datasets, enabling loading of entire codebases with cross-file awareness and synthesis across many long documents like papers or contracts. The change aids building context-aware agents that handle hundreds of tool calls and multi-step workflows. Larger windows reduce repeated retrieval or re-embedding and make multi-day tasks feasible, but they increase compute load, cost, and risk of less focused responses.
Read at InfoQ
Unable to calculate read time
Collection
[
|
...
]