kyutai models
2 kyutai models on Hugging Face, from 2.2B to 2.7B parameters. At the precision each one is published in, the smallest needs about 4.9 GB of VRAM (helium-1-preview-2b, cheapest live fit: RTX 3060) and the largest about 6.0 GB (stt-2.6b-en-trfs, cheapest live fit: RTX 3060). The cheapest way to run helium-1-preview-2b is $0.110/hr.
kyutai models
| Model | Parameters | Published as | VRAM needed | Live GPU fit | Est. $/hr |
|---|---|---|---|---|---|
| helium-1-preview-2b | 2.2B | BF16 | 4.9 GB | RTX 3060 | $0.110/hr |
| stt-2.6b-en-trfs | 2.7B | BF16 | 6.0 GB | RTX 3060 | $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.
