DeepSeek open-sources V4-Flash-Vision-Exp multimodal model

max_paperclips · x · 2026-08-31

DeepSeek released and open-sourced the experimental multimodal model V4-Flash-Vision-Exp. It matches V4-Flash in text capabilities but achieves a major leap in multimodal agent benchmarks, bringing performance close to Opus-4.8. Available via API with support in DeepSeek Harness 0.1.1.

Related event: DeepSeek Quietly Open-Sources V4-Flash-Vision-Exp, Its First Multimodal Model(12 posts)→

Original post →

More from Models

Models channel →