Pick DeepSeek V4 Flash in expert mode
DeepSeek V4 Flash is now one of the models you can select in expert mode, with a context window over a million tokens.
When you open expert mode to choose the model behind a chat, you pick from a short list of Anthropic, Google, and OpenAI models. DeepSeek V4 Flash now sits alongside them.
DeepSeek V4 Flash in the model picker
Select DeepSeek V4 Flash in expert mode and your chat runs on it, the same way any other model there does. It is pinned to a specific snapshot, so the model answering you today is the model answering you next month.
Room for very long inputs
DeepSeek V4 Flash handles just over a million tokens of context and returns up to 64,000 tokens in a response, which gives you space for large document sets and long threads without trimming them down first.
Your own OpenRouter key
If you have connected an OpenRouter key, DeepSeek V4 Flash runs on it and the usage lands on your account rather than ours.
Open expert mode on your next chat and try it against something long.