DeepSeek V4 Flash and DeepSeek V3.2 in DeepMask: text-only reasoning models with tool use, including a 1M-token context window on V4 Flash.
DeepSeek’s models in DeepMask are text-only reasoning models. They read documents but not images, support tool use, and think before they answer. They are a cost-efficient choice for coding, math, and analysis of long text.
DeepSeek V4 Flash
DeepSeek V3.2
About
DeepSeek V4 Flash is DeepSeek’s newest model in DeepMask. It combines a 1M-token context window with a very large output limit of 393K tokens, which makes it useful for long reports, large code changes, and working through big document sets at low cost.
Specifications
Specification
Value
Model provider
DeepSeek
Hosting
EU
Context window
1M tokens
Max output
393K tokens
Image input
No
Tool use
Yes
Reasoning
Built-in thinking
About
DeepSeek V3.2 is the latest release of the V3 line. It is a good fit for coding, math, and structured analysis where answers are short to medium length.
Specifications
Specification
Value
Model provider
DeepSeek
Hosting
EU
Context window
164K tokens
Max output
8K tokens
Image input
No
Tool use
Yes
Reasoning
Built-in thinking
DeepSeek models don’t accept images. To analyze screenshots, charts, or photos, choose a model with image input, such as Sonnet 5.5 or Gemini 3.8 Flash.