DeepSeek V4 Flash
RetiredDeepSeek retired this model on 10 September 2026, so you can't start new chats with it. Use DeepSeek V4.1 Flash instead. Your previous conversations are still available.
Fast and cheap for conversation, hard problems and multi-step work. Not marketed for code — use V4 Pro for that.
- Best for
- Deep reasoning and agent workflows
- Superseded by
- DeepSeek V4.1 Flash
Context window
1,048,576 tokens
≈ 786,000 words · about 3,150 pages
Max output
393,216 tokens
≈ 295,000 words in a single answer
Modalities
- TextInput and output
- ImageNot supported
- AudioNot supported
- VideoNot supported
Reasoning effort
Default: High.
Tools
- Web searchSupported