Dev Tools · 1h ago
Ollama Benchmark on Jetson Nano: Quantization vs Accuracy
A developer tested Ollama models on a Jetson Nano, finding that qwen2.5:3b-instruct achieved 100% accuracy across quantizations, while llama3.2:3b-instruct dropped to 40% at q2_K. Mistral:7b-instruct hit 100% at q4_K_M but varied at other levels. The results highlight the trade-off between model size and accuracy on low-power hardware.
Meridian48 take
The benchmark is narrow—single use case, single device—but offers useful data points for edge AI developers weighing quantization against reliability.
ollamajetson-nano