Finally got Qwen 3.8 Next running on my v100 6gpu setup (TP2 PP3)

A user successfully deployed Qwen 3.8 Next on a v100 6GPU setup, configured with TP2 PP3. Performance metrics show varying token generation speeds across different input lengths, ranging from 1,389 tok/s for 1K tokens to 2,759 tok/s for 131K tokens. The system achieved a maximum speed of 4,679 tok/s at 16,384 (16K) input length. The user expressed satisfaction with the system's stability, thermal management, and low noise levels.

This report is the first to detail Qwen 3.8 Next's performance on a v100 6GPU setup, providing specific token generation speeds across various input lengths, unlike general benchmarks.