Leaderboard
On-device LLM performance rankings powered by Glicko-2
iPad mini 7th Gen
iOSRank
#34
Rating
1,892
±31 RD
Win Rate
88.0%
Conservative Rating
1,830
TG Rating
1,892
PP Rating
1,892
Matches
276
Record
243W – 33L
Models Tested
| Model | TG Median (tok/s) | PP Median (tok/s) | TG Best | PP Best | Runs |
|---|---|---|---|---|---|
| ERNIE-4.5-0.3B-PT-Q4_K_M | 107.87 | 1661.18 | 107.87 | 1661.18 | 1 |
| gemma-3n-E4B-it-Q4_K_M | 93.63 | 2269.06 | 97.46 | 2285.17 | 2 |
| Qwen3-0.6B.Q4_K_M | 61.10 | 99.68 | 61.10 | 99.68 | 1 |
| LFM2.5-1.2B-Instruct-Q4_K_M | 56.87 | 580.59 | 56.87 | 580.59 | 1 |
| ERNIE-4.5-0.3B-PT-F16 | 52.18 | 1718.47 | 52.18 | 1718.47 | 1 |
| gemma-3n-E2B-it-Q4_K_M | 21.53 | 202.93 | 21.84 | 206.97 | 4 |
| gemma-3n-E2B-it-UD-Q4_K_XL | 20.90 | 202.50 | 20.91 | 204.46 | 3 |
| LFM2.5-1.2B-Instruct-BF16 | 19.47 | 645.25 | 19.47 | 645.25 | 1 |
| LFM2.5-1.2B-Thinking-BF16 | 19.42 | 643.64 | 19.42 | 643.64 | 1 |
| Llama-3.2-3B-Instruct-uncensored-Q4_K_M | 19.23 | 177.13 | 19.23 | 177.13 | 1 |
| gemma-2-2b-it-Q6_K | 18.11 | 247.09 | 19.56 | 269.65 | 2 |
| GeumokGPT-Lite.f16 | 18.10 | 46.68 | 18.10 | 46.68 | 1 |
| SmolLM3-Q4_K_M | 16.35 | 150.70 | 16.35 | 150.70 | 1 |
| gemma-3-4b-it-UD-Q4_K_XL | 12.62 | 130.45 | 12.62 | 130.45 | 1 |
| ggml-model-Q4_0 | 10.72 | 14.16 | 10.72 | 14.16 | 1 |
| qwen2.5-3b-instruct-q5_k_m | 9.69 | 14.79 | 9.69 | 14.79 | 1 |
| Llama-3.2-3B-Instruct-Q6_K | 8.78 | 13.94 | 8.78 | 13.94 | 1 |
| gemma-3n-E4B-it-Q4_0 | 8.24 | 12.19 | 8.24 | 12.19 | 1 |
| SecGPT-7B.IQ4_XS | 5.69 | 8.16 | 5.69 | 8.16 | 1 |
| EVA-Qwen2.5-7B-v0.1-Q4_K_L | 3.71 | 7.30 | 3.71 | 7.30 | 1 |
Head-to-Head Record
1–50 of 176 rows
1 / 4
Performance by App Version
ImprovedRegressed