tensorninja / ninfer-4090View on GitHub
From-scratch C++/CUDA engine for Qwen3.8-27B on one RTX 4090: 3,600 tok/s prefill, 149 tok/s MTP decode, full 262K context. Works with any coding agent over OpenAI and Anthropic APIs, best with Caudra. Per-request LoRA adapters, plus OpenAI Decisions and TypeSafe System One on the same weights.
☆16Oct 22, 2026Updated 2 weeks ago

Alternatives and similar repositories for ninfer-4090

Users that are interested in ninfer-4090 are comparing it to the libraries listed below. We may earn a commission when you buy through links labeled 'Ad' on this page.

Sorting:

Are these results useful?