DeepSeekLaunchAI ModelsOctober 17, 2026

DeepSeek releases V4.1 Flash-1017 with stronger coding and agent capabilities

Read original source →huggingface.co

The MIT-licensed update to V4.1 Flash keeps the 552B backbone (8B activated in prefill, 16B in decode) and cuts task execution time and output length. It scores 45 on the Artificial Analysis Intelligence Index, up 6 points, and 77.0 on DeepSWE vs 74.0 for Opus 5.

How this story unfolded

3 days · 6 reports · 22 community posts · 28 of 31 shown

  1. Oct 17
  2. Oct 18
  3. Oct 19
  4. Oct 20

More stories today

Open the live feed