Moondream models
3 Moondream models from vikhyatk on Hugging Face, from 1.9B to 9.3B parameters, published by vikhyatk in F16 and BF16. The smallest official model, moondream1, needs about 4.2 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 Moondream 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 |
|---|---|---|---|---|---|---|---|
| moondream2 | 1.9B | 4.3 GB | 2.2 GB | 1.1 GB | RTX 3070 | $0.088/hr | not published for this architecture |
| moondream3-preview | 9.3B | 20.7 GB | 10.4 GB | 5.2 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 (3)
| Model | Parameters | Published as | VRAM needed | Live GPU fit | Est. $/hr |
|---|---|---|---|---|---|
| moondream1 | 1.9B | F16 | 4.2 GB | RTX 3070 | $0.088/hr |
| moondream2 | 1.9B | BF16 | 4.3 GB | RTX 3070 | $0.088/hr |
| moondream3-preview | 9.3B | BF16 | 20.7 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.