SmolVLM-3B
33%Compact VLM that comfortably runs on a Jetson Orin Nano. The default 'on-device assistant for cameras' pick.
Cross-chip benchmark matrix
Every supported chip, in matched-pair runs from the Fo’c’sle HIL lab. Sortable by any column — click a header. Cells where the chip can’t run this model show Not supported.
| Chip platform | Quant | Latency p50(ms) | Latency p95(ms) | Throughput | Acc. retention(%) | Power(W) | Memory(MB) | Tested by |
|---|---|---|---|---|---|---|---|---|
N NVIDIA Jetson Thor 2070 TOPS · Module | MIXED | 110.8 | 131.7 | 9 FPS | 99.3 | 111.0 | 4,956 | Publisher |
N NVIDIA Jetson AGX Orin 275 TOPS · Module | MIXED | 152.7 | 188.8 | 6.6 FPS | 98.8 | 50.5 | 4,956 | Community |
H Hailo-10H 40 TOPS · M.2 | INT8 | 315.5 | 403.9 | 3.2 FPS | 98.0 | 3.2 | 3,540 | Fo’c’sle HIL |
Apple Neural Engine (M4) 38 TOPS · SoC | MIXED | 342.2 | 421.9 | 2.9 FPS | 99.4 | 7.3 | 4,956 | Community |
Q Qualcomm QCS8550 48 TOPS · SoC | INT8 | 348.8 | 451.5 | 2.9 FPS | 98.4 | 11.8 | 3,540 | Publisher |
Q Snapdragon 8 Gen 3 NPU 45 TOPS · SoC | INT8 | 352.8 | 434.0 | 2.8 FPS | 97.3 | 7.3 | 3,540 | Publisher |
N NVIDIA Jetson Orin Nano 40 TOPS · SoM | INT8 | 422.1 | 535.7 | 2.4 FPS | 98.7 | 13.3 | 3,540 | Fo’c’sle HIL |
H Hailo-8 26 TOPS · M.2 | Not supported | |||||||
Q Qualcomm QCS6490 12 TOPS · SoC | Not supported | |||||||
π Raspberry Pi 5 + Hailo HAT 26 TOPS · HAT | Not supported | |||||||
A Ambarella CV5 16 TOPS · SoC | Not supported | |||||||
A Ambarella CV72 32 TOPS · SoC | Not supported | |||||||
M MediaTek Genio 700 4 TOPS · SoC | Not supported | |||||||
G Google Coral Edge TPU 4 TOPS · USB | Not supported | |||||||
I Intel Movidius Myriad X 4 TOPS · SoC | Not supported | |||||||
A AMD Versal AI Edge VE2302 22 TOPS · SoC | Not supported | |||||||
R Rockchip RK3588 6 TOPS · SoC | Not supported | |||||||
HIL conditions
All numbers measured on Fo’c’sle HIL rigs in Tel Aviv (primary), Munich (secondary), and Pittsburgh (robotics). Single-stream, batch-1, real preprocessing, real downstream consumer. p50/p95 are over 10,000-frame steady-state windows after a 30-second warm-up. Power draw is package power, not wall power. Memory footprint is the resident model + activations footprint at peak — not on-disk.
Submitted publisher numbers are accepted only if they reproduce within ±8% of an HIL-lab matched run on the same chip in the same input mode. Otherwise they live separately under the Discussion tab.