BLIP models
3 BLIP models from Salesforce on Hugging Face, from 470M to 3.9B parameters, published by Salesforce in F32, with a 512-token context. The smallest official model, blip-image-captioning-large, needs about 2.1 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 BLIP 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 |
|---|---|---|---|---|---|---|---|
| blip-image-captioning-large | 470M | 2.1 GB | 0.5 GB | 0.3 GB | RTX 3070 | $0.088/hr | 1.1 GB |
| blip2-opt-2.7b | 3.7B | 16.7 GB | 4.2 GB | 2.1 GB | RTX A5000 | $0.176/hr | 10.0 GB |
| blip2-flan-t5-xl | 3.9B | 17.6 GB | 4.4 GB | 2.2 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 |
|---|---|---|---|---|---|
| blip-image-captioning-large | 470M | F32 | 2.1 GB | RTX 3070 | $0.088/hr |
| blip2-opt-2.7b | 3.7B | F32 | 16.7 GB | RTX A5000 | $0.176/hr |
| blip2-flan-t5-xl | 3.9B | F32 | 17.6 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.
More from Salesforce
- Salesforce (2 models)