bosonai models
2 bosonai models from bosonai on Hugging Face, from 4.7B to 8.9B parameters, published by bosonai in BF16, with a 32K-token context. The smallest official model, higgs-tts-3-4b, needs about 10.4 GB of VRAM at its published precision; the cheapest live fit is RTX 3060 at $0.110/hr.
Pick a size
One row per official bosonai size: the VRAM it needs at each precision, the cheapest GPU that holds it today, and the KV cache for a 32K-token context.
| Model | Parameters | Native VRAM | FP8 VRAM | INT4 VRAM | Live GPU fit (native) | Est. $/hr | KV cache at 32K |
|---|---|---|---|---|---|---|---|
| higgs-tts-3-4b | 4.7B | 10.4 GB | 5.2 GB | 2.6 GB | RTX 3060 | $0.110/hr | 4.5 GB |
| higgs-audio-v3-8b-stt-v2 | 8.9B | 19.9 GB | 10.0 GB | 5.0 GB | RTX A5000 | $0.176/hr | not published for this architecture |
VRAM is weights times a flat 1.2 overhead; the KV cache is a separate per-model figure at 16-bit, shown where the architecture is published. See the methodology.
Official models (2)
| Model | Parameters | Published as | VRAM needed | Live GPU fit | Est. $/hr |
|---|---|---|---|---|---|
| higgs-tts-3-4b | 4.7B | BF16 | 10.4 GB | RTX 3060 | $0.110/hr |
| higgs-audio-v3-8b-stt-v2 | 8.9B | BF16 | 19.9 GB | RTX A5000 | $0.176/hr |
VRAM is for the precision the model is published in: the weights times a flat 1.2 overhead, with the KV cache not included. 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.