Name: DeepSeek-V4-Flash-Vision-Exp
Author: DeepSeek

Evaluation overview

Overall rank

#17

89.7 points · 197 models

Research rank

#13

95.4 points · 191 models

Data coverage

20 items

2 sources · updated through 2026-08-26

Domain performance

Coding#16 / 177

87.7

Research#13 / 191

95.4

Agents#19 / 143

85.7

Vision#34 / 78

75.9

Multimodal · DeepSeek-V4-Flash-Vision-Exp Official Release

35%

1/ 9

Coding · GLM-5.3-Flash Official Release

57.7%

6/ 37

Agents · GLM-5.3-Flash Official Release

75.9%

5/ 23

Reasoning · GLM-5.3-Flash Official Release

1,675pts

9/ 27

Agents · GLM-5.3-Flash Official Release

55.1%

13/ 47

Coding · DeepSeek-V4-Flash-Vision-Exp Official Release

59.3%

6/ 18

Context (input / output): 1.05M / 393K
Release date: 2026-08
Input modalities: Text · Image
Output modalities: Text
Modes: Supports non-thinking and thinking (thinking by default)
Note: Matches DeepSeek-V4-Flash-0731 on text capabilities and adds multimodal capabilities
Exact release date: 2026-08-21
Release status: Experimental API release
Weight status: No separate weights released
Evaluation source: Official 11-benchmark comparison with 31 scores
API model ID: deepseek-v4-flash-vision-exp