Hailo-10H (Pi AI HAT+ 2) Llama-3.2-1B TPS (vendor)
9.89 tokens/sec compiled-model TPS
Fields
id
benchmarks/comm-hailo10h-llama32-1b-tpsMetric
decode_tok_s
Value
9.89
Unit
tokens/sec compiled-model TPS
Model
Llama-3.2-1B-Instruct
Methodology
Hailo compiled-model TPS table via Hailo-Ollama (HailoRT C++ Ollama-compatible server)
Measured by
vendor
Thermal state
unknown
notes
VENDOR-COMPILED model set tops out at <=2B - cannot run 7B. Qwen2.5-1.5B 7.35; Qwen2-1.5B 8.06; Qwen3-1.7B 4.78; Qwen2-VL-2B 7.04.
Notes
VENDOR-COMPILED model set tops out at <=2B - cannot run 7B. Qwen2.5-1.5B 7.35; Qwen2-1.5B 8.06; Qwen3-1.7B 4.78; Qwen2-VL-2B 7.04.