self-hosted/ai
§01·model · /models

Nemotron Elastic 12B (NVFP4)

llmactiveNVIDIA Internal Scientific Research and Development Model License

12B LLM by NVIDIA (Nemotron Elastic) - an 'elastic' checkpoint that nests smaller deployable sizes from one set of weights without retraining. This row tracks the NVFP4-quantized variant, which fits where BF16 runs out of memory (e.g. RTX 5080). NVIDIA Internal Scientific R&D Model License (gated access).

§02·GPUs that run this model
1 total
GPUVRAMSeriesBest speedMin VRAMWorksBenchmarksRecipe
RTX 508016GB501check ↗

benchmarked·~ runs via recipe (not benchmarked)· untested·doesn't fit