DeepSeek opens weights for DeepSeek-V4-Flash-Vision-Exp, adding vision to the V4 Flash line
2026-08-29
DeepSeek's API documentation lists deepseek-v4-flash-vision-exp as a new experimental vision-capable model in the V4 line. Latent Space's AI News recap reports the corresponding weights are now open, with curator commentary noting the release brings DeepSeek to vision parity with Moonshot and GLM.
Significance 3: Extends the open V4 line with a new modality but is not accompanied by independent benchmarks, so it stays below a frontier-topping or axis-moving release.
Implications · machine-drafted, not owner judgment
A vision-capable open DeepSeek variant keeps the open-weights frontier's feature breadth (not just text benchmarks) converging with closed models, reinforcing the case that open weights remain within reach of frontier capability scope, even though this specific release doesn't move the intelligence-index gap directly.
- An independent vision-benchmark comparison of deepseek-v4-flash-vision-exp against Qwen or GLM vision variants
Sources
- primary DeepSeek API news / changelog retrieved 2026-09-01
- curator [AINews] Fal's H3 Max Live breaks the infinite videogen barrier retrieved 2026-09-01