DeepSeek Retires deepseek-chat and deepseek-reasoner Aliases on July 24: Migration Guide

My take: Starting July 24 at 15:59 UTC, any system calling deepseek-chat or deepseek-reasoner will receive errors, not responses. No extension has been announced. If you have production code using those model names, the fix is a one-line change, but it has to happen before that date: update the model parameter to deepseek-v4-pro or deepseek-v4-flash on the same base URL with the same API key.

One detail worth highlighting: deepseek-reasoner maps to V4 Flash, not V4 Pro. If you were using that alias for complex reasoning tasks, you need to evaluate whether Flash is adequate for your use case or whether you should migrate explicitly to Pro. This is not a consequence-free swap; the reasoning capabilities of Flash and Pro are not equivalent.

For engineering teams that depend on DeepSeek models in production, this is non-optional maintenance. The technical migration is trivial; the real risk is ignoring it or assuming behavior will be identical without verifying it against your own use cases.

Does your organization have a process for periodically reviewing model aliases and versions in production systems, or does this kind of breaking change usually go unnoticed until something breaks?