MiniCPM-V models
3 MiniCPM-V models from OpenBMB on Hugging Face, from 1.3B to 8.7B parameters, published by OpenBMB in BF16, with contexts from 40K to 256K tokens. The smallest official model, MiniCPM-V-4.6, needs about 2.9 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 MiniCPM-V 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 |
|---|---|---|---|---|---|---|---|
| MiniCPM-V-4.6 | 1.3B | 2.9 GB | 1.5 GB | 0.7 GB | RTX 3070 | $0.088/hr | 0.38 GB |
| MiniCPM-V-4_5 | 8.7B | 19.4 GB | 9.7 GB | 4.9 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 |
|---|---|---|---|---|---|
| MiniCPM-V-4.6 | 1.3B | BF16 | 2.9 GB | RTX 3070 | $0.088/hr |
| MiniCPM-V-4.6-Thinking | 1.3B | BF16 | 2.9 GB | RTX 3070 | $0.088/hr |
| MiniCPM-V-4_5 | 8.7B | BF16 | 19.4 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.