Leaderboard
On-device LLM performance rankings powered by Glicko-2
iPhone 15
iOSRank
#31
Rating
1,889
±18 RD
Win Rate
87.6%
Conservative Rating
1,853
TG Rating
1,816
PP Rating
1,917
Matches
812
Record
711W – 101L
Models Tested
| Model | TG Median (tok/s) | PP Median (tok/s) | TG Best | PP Best | Runs |
|---|---|---|---|---|---|
| gemma-3-270m-it-F16 | 72.98 | 247.90 | 72.98 | 247.90 | 1 |
| Qwen3-0.6B-Q4_K_M | 54.91 | 880.73 | 54.91 | 880.73 | 1 |
| DeepSeek-R1-Distill-Qwen-1.5B-uncensored.IQ4_XS | 21.50 | 264.56 | 21.50 | 264.56 | 1 |
| llama-3.2-1b-instruct-q8_0 | 21.39 | 44.09 | 31.69 | 55.12 | 3 |
| gemma-3-1b-it.fp16 | 19.78 | 38.56 | 19.78 | 38.56 | 1 |
| Qwen3.5-2B-Q5_K_M | 14.40 | 244.91 | 14.40 | 244.91 | 1 |
| gemma-3-4b-it-UD-Q2_K_XL | 13.90 | 157.43 | 13.90 | 157.43 | 1 |
| Qwen_Qwen3-1.7B-Q5_K_L | 13.36 | 21.92 | 13.36 | 21.92 | 1 |
| qwen2.5-3b-instruct-q5_k_m | 13.00 | 150.76 | 14.50 | 162.05 | 4 |
| Llama-3.2-3B-Instruct-Q4_K_M | 9.82 | 111.31 | 9.82 | 111.31 | 1 |
| Phi-3.5-mini-instruct.Q4_K_M | 9.46 | 102.59 | 12.64 | 122.43 | 4 |
| gemma-2-2b-it-Q6_K | 7.85 | 77.23 | 8.06 | 140.20 | 2 |
| gemma-3-4b-it-Q4_K_M | 7.17 | 71.72 | 7.17 | 71.72 | 1 |
| gemma-3-4b-it.Q2_K | 5.92 | 10.09 | 5.92 | 10.09 | 1 |
| Llama-3.2-1B-Instruct-UD-Q8_K_XL | 5.60 | 12.46 | 5.60 | 12.46 | 1 |
| Qwen3.5-2B-Q4_K_M | 5.36 | 136.98 | 5.36 | 136.98 | 1 |
| Llama-3.2-3B-Instruct-Q6_K | 4.71 | 10.25 | 13.50 | 157.08 | 7 |
Head-to-Head Record
1–50 of 273 rows
1 / 6
Performance by App Version
ImprovedRegressed