Inkling 2 Mini hits 200 tok/s on RTX 5090 via vLLM with native NVFP4

VoidAsuka · x · 2026-11-22

A developer reports running Inkling 2 Mini with vLLM on an RTX 5090, getting 200 tok/s out of the box with native NVFP4 support. First impressions: very smooth, with more concise thinking compared to Qwen models.

Original post →

More from Infra

Infra channel →