Rust-Built QuixiInfer Runs DeepSeek MoE on a Single RTX 5090

QuixiAI released QuixiInfer, an open-source inference engine written from scratch in Rust that runs the 125B-parameter DeepSeek MoE model on a single RTX 5090, achieving 25-30 tok/s at 256k context length.

2026-11-20 ~ 2026-11-20 · 3 related posts