BLOOM models
6 BLOOM models on Hugging Face, from 559M to 176.2B parameters. At the precision each one is published in, the smallest needs about 1.2 GB of VRAM (bloom-560m, cheapest live fit: V100) and the largest about 394 GB (bloom, cheapest live fit: A100). The cheapest way to run bloom-560m is $0.088/hr.
BLOOM models
| Model | Parameters | Published as | VRAM needed | Live GPU fit | Est. $/hr |
|---|---|---|---|---|---|
| bloom-560m | 559M | F16 | 1.2 GB | V100 | $0.088/hr |
| bloomz-560m | 559M | F16 | 1.2 GB | V100 | $0.088/hr |
| bloom-1b1 | 1.1B | F16 | 2.4 GB | V100 | $0.088/hr |
| bloom-1b7 | 1.7B | F16 | 3.8 GB | V100 | $0.088/hr |
| bloom-3b | 3.0B | F16 | 6.7 GB | V100 | $0.088/hr |
| bloom | 176.2B | BF16 | 394 GB | A100 × 5 | $6.06/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.
