Models / granite-3.3

granite-3.3 by ibm-granite

2 models (2B to 8B) · ↓ 421K downloads a month · from $0.10/h on RTX 3060 right now. Each model's versions do the same job; quantized ones (AWQ, FP8, 4-bit) need less GPU memory and run on cheaper machines.

granite-3.3 2B

Version	Runtime	Weights	GPU memory	Cheapest machine now
instructibm-granite/granite-3.3-2b-instruct	vLLM	5.1 GB	8.6 GB	$0.10/h RTX 3060

granite-3.3 8B

Version	Runtime	Weights	GPU memory	Cheapest machine now
Originalibm-granite/granite-3.3-8b-base	vLLM	16 GB	21.9 GB	$0.18/h 2x RTX 3060
instructibm-granite/granite-3.3-8b-instruct	vLLM	16 GB	21.9 GB	$0.18/h 2x RTX 3060

Model details come from Hugging Face and are refreshed daily; prices follow the live offers. GPU memory is GridPup's estimate: the weights, room for an 8,192-token conversation (chat models) and the runtime's own memory. Large models are split across the GPUs of one machine ("4x RTX 4090").

Loading…