DeepSeek opens its eyes: V4-Flash-Vision lands at ~9 images per cent
量子位 · wechat · 2026-08-22
After a four-month wait, DeepSeek released deepseek-v4-flash-vision-exp, a multimodal experimental model with native image input (no image generation). It's live on API at Flash pricing: one image costs at most 384 tokens — roughly ¥0.0012 per image under peak/no-cache rates, about 9 images per cent, far cheaper than a competitor's ¥0.02.
QbitAI's hands-on: text capabilities (agent, reasoning, world knowledge) match V4 Flash, but on multimodal agent benchmarks it jumps significantly, approaching Opus-4.8; basic image recognition feels identical to the beta web model. Paired with DeepSeekHarness (updated the same day with native first-party multimodal support), it handled image-to-3D-modeling in Blender and Canvas front-end demos well — though finger-counting remains a common failure mode. The post also mentions the viral "牛来" model on X, widely speculated to be from Zhipu.
Related event: DeepSeek Launches V4-Flash-Vision-Exp, Closing In on Opus 4.8(29 posts)→
More from Models
- Apollo Research: GLM-5.3 spontaneously reasons about how its actions affect its future self — mimi10v3 · 2026-10-16
- Prediction: Opus 5.5-level models on laptops by 2027, backed by Qwen 27B data — cephaloform · 2026-10-16
- LLM-as-a-Verifier lifts Nemotron 3.5 Super by 9-10 points on agentic benchmarks — simonguozirui · 2026-10-16
- Bonsai 2 ternary models tested: 4B is the sweet spot, but 27B unexpectedly fumbles file writes — cephaloform · 2026-10-16
- Meta's Omnilingual ASR, covering 1,600 languages, named a TIME Best Invention of 2026 — AIatMeta · 2026-10-16
- Apollo Research: GLM-5.3 shows unprecedented 'update-aware reasoning' — repligate · 2026-10-16