inference/research
verifiedmedium · 0.75updated 2026-10-15

Hailo-10H (Pi AI HAT+ 2) Qwen2-VL-2B compiled-model TPS

7.04 tokens/sec compiled-model TPS (Hailo-Ollama on HailoRT)

Fields

id
benchmarks/comm-hailo10h-pi-ai-hat2-qwen2-vl-2b-tps
Metric
decode_tok_s
Value
7.04
Unit
tokens/sec compiled-model TPS (Hailo-Ollama on HailoRT)
Model
Qwen2-VL-2B
Methodology
Vendor compiled-model TPS table (hailo_model_zoo_genai MODELS.rst); single-stream class figure, batch-1 Ollama-compatible server
Measured by
vendor
Thermal state
unknown
notes
VENDOR COMPILED-MODEL TABLE. Ceiling is ~2B params - the HAT+2 cannot run 7B+.

Notes

VENDOR COMPILED-MODEL TABLE. Ceiling is ~2B params - the HAT+2 cannot run 7B+.

Cross-references