DeepSeek releases Harness v0.1.5 with Agent Teams, co-trained with V4.1 Flash

teortaxesTex · x · 2026-09-10

DeepSeek's Harness team shipped v0.1.5, deeply co-trained with the DeepSeek V4.1 Flash model, which was specifically optimized for different Harness configurations. An experimental Agent Teams feature is now open for early trial, integrated with model training. The team pitches an "everything is a plugin" philosophy and is hiring to explore agent capability limits via model-harness co-evolution.

Original post →

More from coding & agent

coding & agent channel →