Betelgeuse runs DeepSeek V4.1 Flash at interactive speed on RTX 3060

A developer reports running DeepSeek V4.1 Flash Q8_XL on a budget RTX 3060 12GB + 16GB RAM PC via Betelgeuse, hitting 19 tok/s decode and 62 tok/s prefill. Cache partitioning and keeping the Engram table on NVMe reportedly ease the prefill bottleneck for large out-of-core MoEs.

• Developer runs Qwen 3.8 Flash Next on an RTX 3060 at 22 tok/s2026-10-31
• llama-modchip runs 284B DeepSeek on 64GB RAM, no GPU2026-11-08
• Free Chinese AI Models Approach OpenAI's Top Cyber Model in Hacking2026-11-06
• Report: closed models dominate price-performance frontier as open share hits low2026-11-06
• Delta Inference Engine v0.93 boosts local DeepSeek-V4.1 runs2026-11-05