Leaderboard
On-device LLM performance rankings powered by Glicko-2
Pixel 6 Pro
AndroidRank
#223
Rating
1,197
±15 RD
Win Rate
20.8%
Conservative Rating
1,167
TG Rating
1,200
PP Rating
1,203
Matches
1,180
Record
245W – 935L
Models Tested
| Model | TG Median (tok/s) | PP Median (tok/s) | TG Best | PP Best | Runs |
|---|---|---|---|---|---|
| DeepSeek-R1-Distill-Qwen-1.5B-Q8_0 | 8.46 | 27.53 | 8.46 | 27.53 | 1 |
| Qwen3-0.6B-Q8_0 | 7.57 | 33.85 | 7.57 | 33.85 | 1 |
| Qwen3.5-0.8B-UD-IQ2_XXS | 7.34 | 35.81 | 7.34 | 35.81 | 1 |
| llama-3.2-1b-instruct-q8_0 | 6.09 | 29.45 | 6.27 | 30.60 | 3 |
| qwen2.5-1.5b-instruct-q8_0 | 6.05 | 28.36 | 6.89 | 29.01 | 3 |
| qwen2.5-3b-instruct-q5_k_m | 4.49 | 9.68 | 4.66 | 11.08 | 2 |
| Ministral-3-3B-Instruct-2512-UD-Q4_K_XL | 4.26 | 19.00 | 4.26 | 19.00 | 1 |
| Phi-3.5-mini-instruct.Q4_K_M | 3.26 | 6.18 | 3.26 | 6.18 | 1 |
| gemma-3n-E4B-it-Q4_K_M | 3.04 | 6.90 | 3.04 | 6.90 | 1 |
| gemma-2-2b-it-Q6_K | 2.37 | 7.18 | 2.37 | 7.18 | 1 |
| Llama-3.2-3B-Instruct-Q6_K | 2.15 | 7.11 | 2.15 | 7.11 | 1 |
| Qwen3.5-4B-UD-Q4_K_XL | 1.84 | 11.97 | 1.84 | 11.97 | 1 |
| deepseek-r1-distill-qwen-7b-q3_k_l | 1.64 | 3.20 | 1.64 | 3.20 | 1 |
Head-to-Head Record
1–50 of 294 rows
1 / 6
Performance by App Version
ImprovedRegressed