DeepSeek releases V4-Flash-Vision-Exp: multimodal agent performance nears Opus-4.8
ying11231 · x · 2026-09-01
DeepSeek has released DeepSeek-V4-Flash-Vision-Exp, an experimental multimodal model now live on its API platform (model='deepseek-v4-flash-vision-exp') with open weights on Hugging Face (MIT license).
- Matches DeepSeek-V4-Flash on text capabilities including agents, reasoning, and world knowledge
- Makes a major leap over V4-Flash on multimodal agent benchmarks, approaching Opus-4.8 per the official claim
- Model card evals: 83.9 on Terminal-Bench 2.1 (rank 3 in size class), 59.3 on Deep-SWE
- DeepSeek Harness 0.1.1 shipped today with out-of-the-box support; the SGLang team is working on support with an updated cookbook
More from Models
- NVIDIA releases Qwen3.8-Flash-Next FP4 local model with new lossless compression format — TheZachMueller · 2026-10-16
- Are the fastest GPU kernels now written by AI? Kernels are hyper-verifiable, argues one dev — fleetwood___ · 2026-10-16
- Scale AI launches muse spark open: 816B-param MoE with 1M context, billed as top US open-weight model — jiayi_pirate · 2026-10-16
- NVIDIA HF listing for Qwen3.8-Flash-Next-FP4-Local leaks unannounced Qwen3.8 model — TheZachMueller · 2026-10-16
- Muse Spark Open, billed as the most capable US open-weight model, lands on Nebius day one — Arindam_1729 · 2026-10-16
- Model at capacity for a straight week drives frustrated dev back to Opus — joonasvirtanen · 2026-10-16