self-hosted/ai
§01·model · /models

Gemma4 31B

multimodalactiveApache-2.0

31B dense LLM by Google (Gemma 4). Apache-2.0.

§02·GPUs that run this model
2 total
GPUVRAMSeriesBest speedMin VRAMWorksBenchmarksRecipe
RTX 509032GB5061.1tokens/s2recipecheck ↗
RTX 309024GB3034.7tokens/s24GB2recipecheck ↗

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