sarvamai models
2 sarvamai models from sarvamai on Hugging Face, from 32.2B to 105.0B parameters, published by sarvamai in F32 and F8_E4M3, with a 128K-token context. The smallest official model, prism, needs about 117 GB of VRAM at its published precision; the cheapest live fit is AMD MI300X at $2.63/hr.
Pick a size
One row per official sarvamai 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 |
|---|---|---|---|---|---|---|---|
| sarvam-30b | 32.2B | 144 GB | 35.9 GB | 18.0 GB | AMD MI300X | $2.63/hr | 0.59 GB |
| prism | 105.0B | 117 GB | – | 58.7 GB | AMD MI300X | $2.63/hr | 1.1 GB |
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 |
|---|---|---|---|---|---|
| sarvam-30b | 32.2B | F32 | 144 GB | AMD MI300X | $2.63/hr |
| prism | 105.0B | F8_E4M3 | 117 GB | AMD MI300X | $2.63/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.