Observed arrival · 2026-10-22
EdgeTokens’ sourced leaderboard for edge-AI speeds
A searchable database compares measured LLM inference speeds across 17 edge-board configurations, with filters for board, model size, and runtime.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- People comparing boards for local LLM inference
- Worth noticing
- Its best-value comparison reports 13.73 tokens per second per $100 for Llama 3.2 1B on the Jetson Orin Nano.
Field notes
The leaderboard exposes filters for board, model size, and runtime, with row-level details such as quantization, runtime version, generation speed, time-to-first-token, and source date where reported. The best-value view calculates tokens per second per $100 for Llama 3.2 1B; the homepage says the Jetson Orin Nano leads that comparison at 13.73. The listed runs span multiple software stacks and source dates.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
✦PrettyNotable craft visible
●ProPolished or operationally mature
◎NicheUnusually specific use
One card from the complete issue