SM120
The compute capability version number for NVIDIA Blackwell-architecture consumer graphics cards (such as the RTX 50 series); the corresponding data-center professional cards use SM100.
Timeline (last 90 days)
SM120 is the compute capability version number of NVIDIA Blackwell architecture consumer graphics cards (such as the RTX 50 series), while data-center professional cards correspond to SM100.
MiniMax officially only supported SM100 professional servers at first; after the uploader added SM120 support for it, speed rose from 45 tokens/s to 51.6 tokens/s.
Currently only vLLM can run the complete Qwen3.8 Flash Next solution on SM120 GPUs; SGLang does not yet support it and crashes during the warm-up phase, while llama.cpp still has an optimization pull request in progress.
SM120 is the Streaming Multiprocessor number in NVIDIA's Blackwell GPU architecture; Ampere is SM80/SM86, Hopper is SM90, and Blackwell is SM100/SM120.
All Facts (3)
SM120 is the compute capability version number of NVIDIA Blackwell architecture consumer graphics cards (such as the RTX 50 series), while data-center professional cards correspond to SM100.
50%UnverifiedMiniMax officially only supported SM100 professional servers at first; after the uploader added SM120 support for it, speed rose from 45 tokens/s to 51.6 tokens/s.
50%UnverifiedSM120 is the Streaming Multiprocessor number in NVIDIA's Blackwell GPU architecture; Ampere is SM80/SM86, Hopper is SM90, and Blackwell is SM100/SM120.
50%