deepseek-v4.1IMAGE
DeepSeek V4.1 Flash 三值量化：解码 11.8→143 tok/s，保留约 92% FP4 智能
三值量化把 510 GB 基础模型压到 142 GB，单并发解码快约 12.1 倍，保留约 92% FP4 智能，作者认为对编程和研究 agent 更实用。
matteo horvath
@buildingmatteo2 天前收录原推发布于 2026.10.27
matteo horvath @buildingmatteo原推发布于 2026.10.27
11.8 → 143 tok/s. DeepSeek-V4.1-Flash-Ternary. 510 GB base → 142 GB pack. ~12.1x faster decode at concurrency 1. ~92% of FP4 intelligence kept. ternary is not free and we lose ~8% vs FP4 for coding and research agents, a model you can run 12x more often beats a slightly smarter one you can't.
x.com/buildingmatteo/status/2114833006825480536采集于 2026.10.27 12:08