Llama 3.2 1B Instruct: every phone chip's best verified run.
| # | Device | SoC | Model | Backend | Prefill t/s | Decode t/s | RAM | App | Tested |
|---|---|---|---|---|---|---|---|---|---|
| 1 | Apple iPhone16,2 | Apple A17 Pro | llama-3.2-1b-instruct-q4_k_m.gguf | llama.cpp · Q4_K_M | 479.4 | 41.3 | 7 GB | 16.163 (1830) | 6 weeks ago |
| 2 | google Pixel 10 Pro XL | Tensor G5 | Llama-3.2-1B-Instruct-Uncensored.i1-Q4_K_M | gguf · Q4_K_M | 108.0 | 11.0 | 15 GB | v3.6.0-beta145 | 2 weeks ago |
← full leaderboard · API: /v1/board · tokforge.ai
