Bonsai Jetson Benchmark 15W: Performance Metrics for LLMs on Edge Hardware

57 benchmark combinations measure the performance of five Bonsai large language models on an NVIDIA Jetson Orin Nano. The dataset includes results for 12 prompt and generation length configurations, with each combination tested over 20 requests. It was created by YuvrajSingh9886 and last updated on May 27, 2026.

Use Cases

• Compare energy efficiency (tok/J) across different model sizes (1.7B, 4B, 8B) based on the described benchmark.
• Analyze the impact of prompt length (256-2048 tokens) on inference performance based on the described sweep.
• Evaluate the trade-off between generation length (128-512 tokens) and power consumption on a 15W device.
• Benchmark quantized (Q1_0) model performance for edge AI applications based on the described model specifications.

Strengths

• Benchmarks five distinct model sizes (1.7B, 4B, 8B) for comparative analysis.
• Tests 12 distinct prompt and generation length configurations for each model.
• Each configuration is tested with 20 requests, which may provide a stable performance average.
• Uses a specific, reproducible hardware and software stack (NVIDIA Jetson Orin Nano, llama.cpp, CUDA).

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Freshness should be verified as the last update timestamp is in the future (2026-05-27).

Provenance

Source

YuvrajSingh9886 via Hugging Face Datasets.

Collection Method

Likely contains performance benchmarks generated through automated testing of language models on specified hardware.

Freshness

Last updated 2026-05-27 17:23:32