microsoft models
6 microsoft models on Hugging Face, from 176M to 4.7B parameters. At the precision each one is published in, the smallest needs about 0.4 GB of VRAM (DialoGPT-small, cheapest live fit: V100) and the largest about 10.6 GB (Mage-VL, cheapest live fit: RTX 5060 Ti). The cheapest way to run DialoGPT-small is $0.088/hr.
microsoft models
| Model | Parameters | Published as | VRAM needed | Live GPU fit | Est. $/hr |
|---|---|---|---|---|---|
| DialoGPT-small | 176M | F16 | 0.4 GB | V100 | $0.088/hr |
| trocr-base-handwritten | 333M | F32 | 1.5 GB | V100 | $0.088/hr |
| trocr-base-printed | 333M | F32 | 1.5 GB | V100 | $0.088/hr |
| kosmos-2-patch14-224 | 1.7B | F32 | 7.4 GB | V100 | $0.088/hr |
| FrogNano-4B-2609 | 4.7B | BF16 | 10.4 GB | RTX 5060 Ti | $0.110/hr |
| Mage-VL | 4.7B | BF16 | 10.6 GB | RTX 5060 Ti | $0.110/hr |
VRAM is for the precision the model is published in, with the same overhead and an 8,192-token context assumed on every page; see the methodology. The fit shown is the lowest-priced single GPU type that holds the model at that precision, or the lowest-priced multi-GPU set (up to 8) when none does.
More from Microsoft
- Florence-2 (5 models)
- Phi-4 (7 models)
- Phi-3.5 (4 models)
- Phi-3 (6 models)
- VibeVoice (4 models)
