Loading…
Loading…
Models / gemma-4
4 models (12B to 31B) · ↓ 26.5M downloads a month · from $0.18/h on Nvidia V100 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.
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|---|---|---|---|
| OBLITERATED (OBLITERATUS)OBLITERATUS/Gemma-4-12B-OBLITERATED | vLLM | 24 GB | 32.4 GB | $0.45/h Nvidia A40 |
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|---|---|---|---|
| Originalgoogle/gemma-4-26B-A4B | vLLM | 52 GB | 61.8 GB | $0.89/h 2x Nvidia A40 |
| itgoogle/gemma-4-26B-A4B-it | vLLM | 52 GB | 61.8 GB | $0.89/h 2x Nvidia A40 |
| it FP8 dynamic (RedHatAI)RedHatAI/gemma-4-26B-A4B-it-FP8-dynamic | vLLM | 29 GB | 36.3 GB | $0.45/h Nvidia A40 |
| NVFP4 (nvidia)nvidia/Gemma-4-26B-A4B-NVFP4 | vLLM | 19 GB | 25.3 GB | $0.34/h 2x RTX 5060 Ti |
| it AWQ 4bit (cyankiwi)cyankiwi/gemma-4-26B-A4B-it-AWQ-4bit | vLLM | 17 GB | 23.6 GB | $0.24/h RTX 3090 |
| it NVFP4 (RedHatAI)RedHatAI/gemma-4-26B-A4B-it-NVFP4 | vLLM | 16 GB | 22.7 GB | $0.24/h RTX 3090 |
| it NVFP4 (unsloth)unsloth/gemma-4-26B-A4B-it-NVFP4 | vLLM | 17 GB | 23.2 GB | $0.24/h RTX 3090 |
| it NVFP4 (bg-digitalservices)bg-digitalservices/Gemma-4-26B-A4B-it-NVFP4 | vLLM | 16 GB | 22.7 GB | $0.24/h RTX 3090 |
| it fp8 (LargitData)LargitData/gemma-4-26b-a4b-it-fp8 | vLLM | 27 GB | 34.6 GB | $0.45/h Nvidia A40 |
| it NVFP4 (cklaus)cklaus/gemma-4-26B-A4B-it-NVFP4 | vLLM | 18 GB | 24.5 GB | $0.34/h 2x RTX 5060 Ti |
| it gpuep 1.0 gpu 128 (amd)amd/Gemma-4-26B-A4B-it_gpuep_1.0_gpu_128 | ONNX Runtime GenAI | 17 GB | 22.6 GB | $0.18/h Nvidia V100 |
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|---|---|---|---|
| gemma 4 26B A4B it qat AWQ INT4 (cyankiwi)cyankiwi/gemma-4-26B-A4B-it-qat-AWQ-INT4 | vLLM | 17 GB | 23.6 GB | $0.24/h RTX 3090 |
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|---|---|---|---|
| Originalgoogle/gemma-4-31B | vLLM | 63 GB | 80.7 GB | $0.89/h 2x Nvidia A40 |
| itgoogle/gemma-4-31B-it | vLLM | 63 GB | 80.7 GB | $0.89/h 2x Nvidia A40 |
| it qat w4a16 ctgoogle/gemma-4-31B-it-qat-w4a16-ct | vLLM | 23 GB | 37 GB | $0.45/h Nvidia A40 |
| it qat q4 0 unquantizedgoogle/gemma-4-31B-it-qat-q4_0-unquantized | vLLM | 63 GB | 80.7 GB | $0.89/h 2x Nvidia A40 |
| IT NVFP4 (nvidia)nvidia/Gemma-4-31B-IT-NVFP4 | vLLM | 33 GB | 47.4 GB | $0.46/h RTX A6000 |
| it FP8 block (RedHatAI)RedHatAI/gemma-4-31B-it-FP8-block | vLLM | 33 GB | 48.1 GB | $0.78/h 4x RTX 4060 Ti |
| it qat NVFP4 Blackwell (melcheikh)melcheikh/gemma-4-31B-it-qat-NVFP4-Blackwell | vLLM | 20 GB | 33.9 GB | $0.45/h Nvidia A40 |
| it AWQ (QuantTrio)QuantTrio/gemma-4-31B-it-AWQ | vLLM | 20 GB | 33.9 GB | $0.45/h Nvidia A40 |
| it FP8 dynamic (RedHatAI)RedHatAI/gemma-4-31B-it-FP8-dynamic | vLLM | 33 GB | 48.1 GB | $0.78/h 4x RTX 4060 Ti |
| it NVFP4 turbo (LilaRest)LilaRest/gemma-4-31B-it-NVFP4-turbo | vLLM | 19 GB | 32.6 GB | $0.45/h Nvidia A40 |
| it AWQ 4bit (cyankiwi)cyankiwi/gemma-4-31B-it-AWQ-4bit | vLLM | 21 GB | 34.4 GB | $0.45/h Nvidia A40 |
| it 4bit W4A16 AWQ (ebircak)ebircak/gemma-4-31B-it-4bit-W4A16-AWQ | vLLM | 19 GB | 32.5 GB | $0.45/h Nvidia A40 |
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").