Leaderboard
On-device LLM performance rankings powered by Glicko-2
iPhone 15 Plus
iOSRank
#19
Rating
1,929
±18 RD
Win Rate
91.4%
Conservative Rating
1,892
TG Rating
1,874
PP Rating
1,936
Matches
801
Record
732W – 69L
Models Tested
| Model | TG Median (tok/s) | PP Median (tok/s) | TG Best | PP Best | Runs |
|---|---|---|---|---|---|
| gemma-3-270m-it-IQ4_NL | 109.41 | 326.48 | 109.41 | 326.48 | 1 |
| DeepSeek-R1-Distill-Qwen-1.5B-Q4_0 | 36.58 | 421.66 | 36.58 | 421.66 | 1 |
| llama-3.2-1b-instruct-q8_0 | 30.30 | 581.26 | 30.30 | 581.26 | 1 |
| deepseek-ai.DeepSeek-R1-Distill-Qwen-1.5B.Q8_0 | 24.42 | 416.45 | 24.42 | 416.45 | 1 |
| SmolLM2-1.7B-Instruct-Q8_0 | 21.18 | 329.31 | 21.21 | 338.43 | 2 |
| google_gemma-3-1b-it-bf16 | 20.81 | 693.55 | 20.84 | 697.69 | 3 |
| DeepSeek-R1-Distill-Qwen-1.5B-Q4_K_M | 20.78 | 31.78 | 20.78 | 31.78 | 1 |
| gemma-3-1b-it.fp16 | 20.28 | 39.46 | 20.28 | 39.46 | 1 |
| Gemmasutra-Mini-2B-v1-Q6_K | 16.93 | 226.93 | 16.93 | 226.93 | 1 |
| gemma-2-2b-it-Q6_K | 15.06 | 211.17 | 18.66 | 250.44 | 3 |
| Vikhr-Gemma-2B-instruct-Q4_1 | 13.96 | 23.07 | 14.36 | 23.28 | 3 |
| Falcon3-3B-Instruct-abliterated-Q6_K | 13.87 | 162.75 | 13.87 | 162.75 | 1 |
| Hermes-3-Llama-3.2-3B.Q6_K | 12.62 | 147.97 | 12.62 | 147.97 | 1 |
| DeepSeek-R1-Distill-Qwen-1.5B-IQ2_M | 12.62 | 16.17 | 12.62 | 16.17 | 1 |
| blacksheep-llama3.2-3b-q6_k | 11.01 | 135.43 | 11.01 | 135.43 | 1 |
| Dolphin3.0-Qwen2.5-3b-Q6_K | 8.98 | 107.47 | 8.98 | 107.47 | 1 |
| DeepSeek-R1-Distill-Qwen-1.5B-Abliterated-dpo.Q6_K | 8.62 | 16.00 | 8.62 | 16.00 | 1 |
| SmallThinker-3B-Preview-Q5_K_L | 8.56 | 68.37 | 11.33 | 127.02 | 2 |
| qwen2.5-3b-instruct-q5_k_m | 8.21 | 11.66 | 8.21 | 11.66 | 1 |
| Llama-3.2-3B-Instruct-Q6_K | 7.81 | 116.56 | 7.81 | 116.56 | 1 |
| Dolphin3.0-Llama3.2-3B-Q6_K | 6.52 | 94.36 | 6.52 | 94.36 | 1 |
| Phi-3.5-mini-instruct.Q4_K_M | 4.20 | 6.26 | 6.54 | 7.38 | 2 |
Head-to-Head Record
1–50 of 301 rows
1 / 7
Performance by App Version
ImprovedRegressed