AnalysisAI ModelsDevelopers
Jun 2, 5:52 PM
Bringing up DeepSeek-V4-Flash on AMD MI300X
A worklog details the process of running DeepSeek-V4-Flash with vLLM on AMD MI300X GPUs, covering FP8 precision and software compatibility issues. The author notes that as of May 2026, vLLM with DeepSeek-V4-Flash does not work on MI300X out of the box.