deepseek-v4.1POST
V4.1 Flash 1017 跑分:Terminal-Bench 4.0 升至 42.3
官方数据:Terminal-Bench 4.0 从 31.2 升到 42.3,DeepSWE 74.2→77.0,Automation-Bench 54.8→60.5,输出更短、更快更便宜,MIT 许可权重上 HF。
Neo
@NeoAIForecast10 天前收录原推发布于 2026.10.17
Neo @NeoAIForecast原推发布于 2026.10.17
DeepSeek is back baby!!! 🐋 Deepseek-V4.1-Flash-1017 Terminal-Bench 4.0: 31.2 > 42.3 DeepSWE v1.1: 74.2 > 77.0 Automation-Bench: 54.8 > 60.5 Shorter outputs. Faster tasks. Lower cost per task, per DeepSeek. MIT license. Weights on Hugging Face today. https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash-1017
x.com/neoaiforecast/status/2111397629615886468采集于 2026.10.18 16:42