Leaderboard
On-device LLM performance rankings powered by Glicko-2
iPhone 13 Mini
iOSRank
#76
Rating
1,864
±89 RD
Win Rate
87.5%
Conservative Rating
1,687
TG Rating
1,834
PP Rating
1,864
Matches
32
Record
28W – 4L
Models Tested
| Model | TG Median (tok/s) | PP Median (tok/s) | TG Best | PP Best | Runs |
|---|---|---|---|---|---|
| gemma-3-270m-it-qat-Q4_0 | 106.13 | 417.51 | 106.13 | 417.51 | 1 |
| gemma-3-270m-it-UD-Q8_K_XL | 62.78 | 405.60 | 62.78 | 405.60 | 1 |
| qwen2.5-0.5b-instruct-q6_k | 48.16 | 108.84 | 48.16 | 108.84 | 1 |
| gemma-3-1b-it-Q4_K_M | 28.64 | 512.04 | 28.64 | 512.04 | 1 |
| DeepSeek-R1-Distill-Qwen-1.5B-Q4_1 | 24.57 | 360.91 | 25.04 | 363.51 | 2 |
| llama-3.2-1b-instruct-q8_0 | 20.05 | 45.77 | 22.60 | 511.28 | 3 |
| totob-1.5B | 19.76 | 288.22 | 19.76 | 288.22 | 1 |
| DeepSeek-R1-Distill-Qwen-1.5B-Q2_K | 17.40 | 23.80 | 17.40 | 23.80 | 1 |
| minithinky-v2-1b-llama-3.2-q8_0 | 16.86 | 403.60 | 16.86 | 403.60 | 1 |
| SmolLM2-1.7B-Instruct-Q4_K_L | 13.84 | 151.70 | 13.84 | 151.70 | 1 |
| DeepSeek-R1-Distill-Qwen-1.5B-IQ2_M | 11.99 | 190.78 | 11.99 | 190.78 | 1 |
| granite-3.2-2b-instruct-Q3_K_L | 10.23 | 105.07 | 10.23 | 105.07 | 1 |
| Llama-3.2-3B-Instruct-Q3_K_L | 2.17 | 5.04 | 2.17 | 5.04 | 1 |
| DeepSeek-R1-Distill-Qwen-1.5B-Q8_0 | 0.41 | 25.64 | 0.41 | 25.64 | 1 |
| Phi-3.5-mini-instruct.Q4_K_M | 0.20 | 7.52 | 0.20 | 7.52 | 1 |
Head-to-Head Record
1–50 of 127 rows
1 / 3
Performance by App Version
ImprovedRegressed