Basalt 0.2.0 speeds Qwen 3.8 Flash Next on Blackwell GPUs

Basalt is a Blackwell-only inference engine for Qwen 3.8 Flash Next in NVFP4. Version 0.2.0 reports about 300 tokens/s prose, 600 tokens/s structured output and 10K tokens/s prefill on a 5090 plus 5060 Ti.

Domain

general

Read it at Reddit →

Coverage: 1 sources, in order

1. 03 Nov 20:49 Reddit Read the article →

← Front page