Leaderboard
On-device LLM performance rankings powered by Glicko-2
Pixel 7 Pro
AndroidRank
#168
Rating
1,390
±15 RD
Win Rate
39.4%
Conservative Rating
1,361
TG Rating
1,387
PP Rating
1,420
Matches
1,240
Record
489W – 751L
Models Tested
| Model | TG Median (tok/s) | PP Median (tok/s) | TG Best | PP Best | Runs |
|---|---|---|---|---|---|
| gemma-3-270m-it-F16 | 14.44 | 116.89 | 14.44 | 116.89 | 1 |
| llama-3.2-1b-instruct-q8_0 | 9.75 | 47.33 | 11.45 | 57.35 | 2 |
| DeepSeek-R1-Distill-Qwen-1.5B-IQ2_M | 7.32 | 13.44 | 7.32 | 13.44 | 1 |
| qwen2.5-coder-3b-instruct-q4_k_m | 6.81 | 17.65 | 6.81 | 17.65 | 1 |
| SmolLM2-1.7B-Instruct-Q8_0 | 5.93 | 24.08 | 5.93 | 24.08 | 1 |
| SmallThinker-3B-Preview-Q4_K_M | 5.38 | 13.47 | 5.38 | 13.47 | 1 |
| Qwen3-1.7B-Q8_0 | 5.04 | 26.15 | 5.04 | 26.15 | 1 |
| gemma-3-1b-it.Q5_K_M | 4.46 | 30.32 | 4.46 | 30.32 | 1 |
| Gemmasutra-Mini-2B-v1-Q6_K | 4.14 | 13.01 | 4.14 | 13.01 | 1 |
| qwen2.5-3b-instruct-q5_k_m | 3.66 | 10.11 | 3.66 | 10.11 | 1 |
| Llama-3.2-3B-Instruct-Q6_K | 3.61 | 10.01 | 3.91 | 11.38 | 5 |
| Phi-4-mini-instruct.Q4_K_M | 3.36 | 9.06 | 3.36 | 9.06 | 1 |
| Phi-3.5-mini-instruct.Q4_K_M | 3.34 | 8.61 | 4.34 | 10.09 | 4 |
| gemma-2-2b-it-Q6_K | 3.04 | 10.43 | 5.56 | 16.62 | 6 |
| Qwen3-4B.Q6_K | 2.62 | 6.26 | 3.11 | 8.29 | 2 |
| DeepSeek-R1-Distill-Qwen-7B-Q6_K | 2.43 | 5.08 | 2.43 | 5.08 | 1 |
| gemma-3-4B-it-QAT-Q4_0 | 2.26 | 11.27 | 2.26 | 11.27 | 1 |
Head-to-Head Record
1–50 of 305 rows
1 / 7
Performance by App Version
ImprovedRegressed