DeepSeek launches V4-Flash-Vision-Exp, multimodal agent performance nears Opus-4.8

OwariDa · x · 2026-08-22

DeepSeek officially launched the experimental multimodal model DeepSeek-V4-Flash-Vision-Exp on its API platform: text capabilities (agents, reasoning, world knowledge) match V4-Flash, while multimodal agent benchmarks take a major leap, approaching Opus-4.8. Available via model='deepseek-v4-flash-vision-exp'; the simultaneously released DeepSeek Harness 0.1.1 supports it out of the box. The retweeter jokes that it being "strictly stronger than 0731" was because it was a blinded VLM all along.

Related event: DeepSeek Launches V4-Flash-Vision-Exp, Closing In on Opus 4.8 for Multimodal Agents(27 posts)→

Original post →

More from Fun

Fun channel →