๐Ÿ“ฑ Updated benchmarks: 9 models tested on Snapdragon 865

#1
by 3morixd - opened
dispatchAI org

We've updated our mobile AI leaderboard with fresh benchmarks from the phone farm.

Latest results:

  • SmolLM2-135M: 22.8 t/s (fastest!)
  • Qwen2.5-0.5B: 19.2 t/s
  • TinyLlama-1.1B: 18.5 t/s
  • Qwen2.5-1.5B: 16.9 t/s
  • Llama-3.2-1B: 16.3 t/s

All tested on 40 Samsung S20 FE (Snapdragon 865, 8GB RAM) with Q4_K_M GGUF.

Which model should we benchmark next? Comment below!

  • Dispatch AI (FZE), Sharjah UAE

Sign up or log in to comment