sm120
DeepSeek-v4.1-Flash-EXL3-2.0bpw-Ablit-EngramQ4-SM120-Dual-RTX-Pro-6000MiniMax-M3-sm120-safe-GGUFflux2dev-fp8-transformer-rtx5090-sm120Flash_attn2_2.8.3_cu128_sm120_cp312_cu128_torch210_wheelflash-attn-4-sm120DSv4-Flash-FP8-SM120-Configstriattention-sm120-verificationFlux2Transformer2DModel-sm120-cu130-r77
Datasets
All datasets matching “sm120”sm120-field-guide
the sm_120 field guide
AI on consumer Blackwell. the fixes and footguns from one RTX 5090, measured honestly.
every entry here is something I hit running modern AI on an RTX 50-series card: the error message, the mechanism behind it, the fix that worked, and the date + versions it was verified on. if I didn't run it, it isn't here. if you own a 50-series card (sm_120), this is the debugging session you get to skip.
most AI infrastructure targets datacenter cards. consumer… See the full description on the dataset page: https://huggingface.co/datasets/witcheer/sm120-field-guide.blackwell-sm120-serving-matrix
Blackwell SM120 Serving Matrix
Measured serving results for LLMs on workstation and server Blackwell (sm_120, RTX PRO 6000, 96 GB), across vLLM, SGLang and llama.cpp, with FP8 and NVFP4 passes. Each row is one exact cell: a model, a runtime and version, a quantization, a topology and a load point, with the startup result, throughput or latency where measured, and a link to the raw artifact.
The point of the matrix is the cells that are known-broken as much as the ones that are… See the full description on the dataset page: https://huggingface.co/datasets/jahnclawdmonet/blackwell-sm120-serving-matrix.