Jamba models
3 Jamba models from ai21labs on Hugging Face, from 3.0B to 51.6B parameters, published by ai21labs in BF16, with a 256K-token context. The smallest official model, AI21-Jamba2-3B, needs about 6.8 GB of VRAM at its published precision; the cheapest live fit is RTX 3070 at $0.088/hr.
Pick a size
One row per official Jamba 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 |
|---|---|---|---|---|---|---|---|
| AI21-Jamba2-3B | 3.0B | 6.8 GB | 3.4 GB | 1.7 GB | RTX 3070 | $0.088/hr | 0.03 GB |
| AI21-Jamba-Reasoning-3B | 3.2B | 7.1 GB | 3.6 GB | 1.8 GB | RTX 3070 | $0.088/hr | 0.03 GB |
| Jamba-v0.1 | 51.6B | 115 GB | 57.6 GB | 28.8 GB | AMD MI300X | $2.63/hr | 0.50 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 (3)
| Model | Parameters | Published as | VRAM needed | Live GPU fit | Est. $/hr |
|---|---|---|---|---|---|
| AI21-Jamba2-3B | 3.0B | BF16 | 6.8 GB | RTX 3070 | $0.088/hr |
| AI21-Jamba-Reasoning-3B | 3.2B | BF16 | 7.1 GB | RTX 3070 | $0.088/hr |
| Jamba-v0.1 | 51.6B | BF16 | 115 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.