Leaderboard
On-device LLM performance rankings powered by Glicko-2
Pixel 10 Pro
AndroidRank
#154
Rating
1,445
±17 RD
Win Rate
44.7%
Conservative Rating
1,410
TG Rating
1,378
PP Rating
1,739
Matches
911
Record
407W – 504L
Models Tested
| Model | TG Median (tok/s) | PP Median (tok/s) | TG Best | PP Best | Runs |
|---|---|---|---|---|---|
| LFM2.5-1.2B-Instruct-Q4_K_M | 29.72 | 198.87 | 29.72 | 198.87 | 1 |
| LFM2.5-1.2B-Thinking-Q4_K_M | 27.61 | 199.31 | 27.61 | 199.31 | 1 |
| LFM2-2.6B-Exp-Q4_K_M | 12.27 | 87.07 | 12.27 | 87.07 | 1 |
| Qwen3.5-2B-Q4_K_M | 10.35 | 111.60 | 10.35 | 111.60 | 1 |
| SmolLM2-1.7B-Instruct-Q8_0 | 6.22 | 71.66 | 6.22 | 71.66 | 1 |
| gemma-3-1b-it.Q2_K | 5.24 | 74.93 | 5.24 | 74.93 | 1 |
| Llama-3.2-3B-Instruct-Q6_K | 5.12 | 25.18 | 5.12 | 25.18 | 1 |
| Dolphin-X1-8B-q4_k_m | 4.92 | 30.74 | 4.92 | 30.74 | 1 |
| Deepseek-R1-0528-Qwen3-8B-Q4_K_M | 4.80 | 24.14 | 4.80 | 24.14 | 1 |
| Qwen3.5-4B-Q4_K_M | 4.71 | 39.21 | 4.71 | 39.21 | 1 |
| gemma-2-2b-it-Q6_K | 4.64 | 34.52 | 6.46 | 40.70 | 2 |
| Phi-3.5-mini-instruct.Q4_K_M | 3.50 | 19.91 | 3.50 | 19.91 | 1 |
| qwen2.5-3b-instruct-q5_k_m | 1.83 | 17.67 | 2.29 | 20.46 | 2 |
| ARM-DS-R1-Qwen3-8B-ArliAI-RpR-v4-Small-Q4_0-imat | 1.49 | 18.27 | 1.65 | 19.09 | 2 |
| Qwen3.5-9B-Q4_K_M | 1.39 | 14.31 | 1.39 | 14.31 | 1 |
Head-to-Head Record
1–50 of 239 rows
1 / 5
Performance by App Version
ImprovedRegressed