SemiAnalysis says Nvidia vLLM ran DeepSeekv4.1 Flash at launch while AMD's did not
SemiAnalysis said on September 11, 2026 that AMD's vLLM documentation pointed to a software image that had not been publicly released 23 hours after the model's launch.
TLDR
SemiAnalysis reported on September 11, 2026 that DeepSeekv4.1 Flash worked out of the box with no issues on Nvidia vLLM across six SKUs: H100, H200, B200, B300, GB200 and GB300. By contrast, it said AMD vLLM still did not work with the model 23 hours after release. AMD's documentation pointed to the image vllm/vllm-openai-rocm:deepseekv41-flash-0909, which SemiAnalysis said AMD had not yet publicly released. The publisher called for AMD to improve its process for supporting models at launch.
Combined views
13.1K
1 Source, first seen 19d ago