DeepSeek releases DeepSeek-V4.1-Flash-1017 with improved reasoning, coding, and vision

The official model card describes DeepSeek-V4.1-Flash-1017 as a 763B model with a 1M-token context, with weights and repository code under MIT. DeepSeek reports improvements from 31.2 to 42.3 on Terminal-Bench 4.0 and from 74.2 to 77.0 on DeepSWE v1.1, alongside shorter execution times and outputs.

Visit original source