deepseek-v4.1POST
DeepSeek V4.1 Flash 1017 多项基准提升，Terminal-Bench 涨 11.1 分
官方图表显示 Terminal-Bench 4.0 从 31.2% 升至 42.3%，DeepSWE 77.0%，价格不变、输出更短，MIT 开源权重已上线。
Neelesh
@gadi_neelesh10 天前收录原推发布于 2026.10.17
Neelesh @gadi_neelesh原推发布于 2026.10.17
🤯 DeepSeek V4.1 Flash just gained 11.1 percentage points on Terminal-Bench 4.0 in a "minor update." The new 1017 version brings some serious benchmark jumps: Terminal-Bench 4.0: 31.2% → 42.3% ProgramBench: 20.3% → 29.1% DeepSWE: 74.2% → 77.0% Automation-Bench: 54.8% → 60.5% It even beats GPT-5.6 Sol on Terminal-Bench in DeepSeek's tests. And the crazy part? Same API pricing, shorter outputs and reportedly faster task completion. Open weights. MIT license. Already live.
这条是引用着另一条写的：x.com/deepseek_ai/status/2111375640717033777
x.com/gadi_neelesh/status/2111386441880424924采集于 2026.10.18 16:45