VLM Inference Benchmark Explorer
MiniCPM-V-4.7-1B on DGX Spark — BF16, vLLM vision-language inference benchmark
Open this configuration in the VLM Inference Benchmark Explorer
| Model | MiniCPM-V-4.7-1B |
|---|---|
| Parameters | 1.6B |
| Device | DGX Spark |
| Quantization | BF16 |
| Image size | 720p |
| Images per camera | 1 |
| Max Cameras (response ≤ 3 s) | 64 |
| Engine | vLLM |
| Cameras | Response (s) | TPS (tok/s) | Target |
|---|---|---|---|
| 1 | 0.20 | 115.6 | ≤3 s |
| 2 | 0.26 | 116.8 | ≤3 s |
| 4 | 0.36 | 106.1 | ≤3 s |
| 8 | 0.58 | 88.4 | ≤3 s |
| 16 | 1.06 | 69.4 | ≤3 s |
| 32 | 2.02 | 46.6 | ≤3 s |
| 64 | 2.86 | 24.2 | ≤3 s |
| 128 | 4.54 | 12.5 | >3 s |
| 256 | 8.96 | 6.8 | >3 s |
| 320 | 14.6 | 6.0 | >3 s |
| 352 | 17.7 | 5.9 | >3 s |
| 384 | — | — | Failed |
| Images per request | Response (s) | TPS (tok/s) | Target |
|---|---|---|---|
| 1 | 0.20 | 115.6 | ≤3 s |
| 2 | 0.32 | 115.4 | ≤3 s |
| 4 | 0.60 | 114.4 | ≤3 s |
| 8 | 1.18 | 113.1 | ≤3 s |
| 16 | 2.34 | 110.9 | ≤3 s |
| 32 | 4.71 | 106.4 | >3 s |
| 64 | 9.68 | 97.6 | >3 s |
| 128 | 20.0 | 84.0 | >3 s |
| 256 | — | — | Failed |
MiniCPM-V-4.7-1B (1.6B parameters), served in BF16 format with vLLM on DGX Spark, starts answering a single camera that sends one 720p image after 0.2 s, then writes 115.6 tok/s.
With 128 images per camera at 720p, the answer starts after 20.02 s. With one 2K image, after 0.29 s.
At the default target of an answer starting within 3 s (720p, one image per camera), it keeps up with 64 cameras at once.
MiniCPM-V-4.7-1B (1.6B parameters), served in BF16 format with vLLM on DGX Spark, starts answering a single camera that sends one 720p image after 0.2 s, then writes 115.6 tok/s.
With 128 images per camera at 720p, the answer starts after 20.02 s. With one 2K image, after 0.29 s.
At the default target of an answer starting within 3 s (720p, one image per camera), it keeps up with 64 cameras at once.