Model comparison
DeepSeek V4 Flash 0731 vs DeepSeek V3
Capability radar, value bars, and side-by-side posture.
DeepSeek V4 Flash 0731
Not scored
value score
DeepSeek V3
87
value score
DeepSeek V4 Flash 0731 ctx
—
DeepSeek V3 ctx
128K
Capability radar
Value · context · multimodal · openness · speed posture
Value head-to-head
- DeepSeek V387
Attribute tape
| Field | DeepSeek V4 Flash 0731 | DeepSeek V3 |
|---|---|---|
| Developer | ~deepseek | ~deepseek |
| Context | — | 128K |
| Modalities | Text, Code | |
| Openness | — | Open Weights |
| Speed | — | — |
| Price | — | $0.27/1M in · $1.1/1M out |
| Value score | — | 87 |
| Summary | DeepSeek V4 Flash 0731 is the official public-beta release of DeepSeek's V4-Flash model: a 284B-parameter mixture-of-experts (MoE) LLM with only 13B active parameters, re-post-trained from the earlier V4-Flash preview for improved performance on agentic, coding, and tool-use tasks. It is accessible via DeepSeek's own API and third-party inference providers including OpenRouter, DeepInfra, and Fireworks. | Strong open-weights MoE general model. |