Running DeepSeek V4.1 Flash Locally: 27 tok/s on an RTX 5090

Developers demonstrated local deployments of DeepSeek V4.1 Flash, with one achieving 15 tok/s on a $2,500 RTX 3090 rig and QuixiAI reaching 27 tok/s at 256k context on an RTX 5090 using its open-source Rust-based inference engine QuixiInfer.

2026-11-08 ~ 2026-11-08 · 3 related posts

• Running DeepSeek V4.1 Flash locally at 15 tok/s on an RTX 3090 build for ~$2.5k — casper_hansen_ · 2026-11-08
• DeepSeek v4.1 Flash runs locally at 27 tok/s decode with 256k context on an RTX 5090 — QuixiAI · 2026-11-08
• Single RTX 5090 serves 125B MoE at 256k context: QuixiInfer hits 27 tok/s decode — QuixiAI · 2026-11-08