Leaderboard
On-device LLM performance rankings powered by Glicko-2
Galaxy S23
AndroidRank
#95
Rating
1,643
±14 RD
Win Rate
63.8%
Conservative Rating
1,615
TG Rating
1,661
PP Rating
1,531
Matches
1,367
Record
872W – 495L
Models Tested
| Model | TG Median (tok/s) | PP Median (tok/s) | TG Best | PP Best | Runs |
|---|---|---|---|---|---|
| gemma-3-270m-it-UD-Q6_K_XL | 29.92 | 295.79 | 29.92 | 295.79 | 1 |
| Qwen_Qwen3-0.6B-Q4_K_M | 24.46 | 87.34 | 24.46 | 87.34 | 1 |
| Qwen3-0.6B.Q4_K_M | 18.71 | 59.70 | 18.71 | 59.70 | 1 |
| Qwen3-0.6B.Q5_K_M | 16.57 | 54.71 | 16.57 | 54.71 | 1 |
| Qwen3-0.6B.Q2_K | 15.75 | 57.48 | 15.75 | 57.48 | 1 |
| llama-3.2-1b-instruct-q8_0 | 15.19 | 65.80 | 20.44 | 81.56 | 8 |
| qwen2.5-coder-1.5b-q8_0 | 14.07 | 44.59 | 14.07 | 44.59 | 1 |
| Qwen3-0.6B.Q6_K | 12.97 | 28.84 | 16.10 | 52.14 | 2 |
| gemma-3-1b-it.Q3_K_L | 12.61 | 62.51 | 12.61 | 62.51 | 1 |
| qwen2.5-1.5b-instruct-q8_0 | 12.54 | 40.81 | 13.82 | 49.24 | 6 |
| DeepSeek-R1-Distill-Qwen-1.5B-Q5_K_M | 12.52 | 29.40 | 12.55 | 31.39 | 2 |
| SmolLM2-1.7B-Instruct-Q8_0 | 12.46 | 37.11 | 12.46 | 37.11 | 1 |
| Qwen3-0.6B.Q3_K_M | 12.26 | 50.17 | 12.26 | 50.17 | 1 |
| gemma-3-1b-it-Q4_0 | 11.94 | 224.87 | 11.94 | 224.87 | 1 |
| Qwen3-1.7B-Q3_K_S | 11.06 | 25.11 | 11.06 | 25.11 | 1 |
| Qwen3-1.7B.Q2_K | 10.41 | 20.93 | 10.41 | 20.93 | 1 |
| Qwen3-0.6B.fp16 | 10.08 | 25.96 | 10.08 | 25.96 | 1 |
| Llama-3.2-1B-Instruct.IQ1_M | 9.97 | 22.30 | 9.97 | 22.30 | 1 |
| Hermes-3-Llama-3.2-3B.Q4_K_M | 8.69 | 15.14 | 8.69 | 15.14 | 1 |
| Qwen3-4B.Q4_K_M | 8.28 | 12.53 | 8.28 | 12.53 | 1 |
| DeepSeek-R1-Distill-Qwen-1.5B-IQ4_NL | 8.19 | 32.66 | 8.19 | 32.66 | 1 |
| gemma-3-4b-it-uncensored.i1-Q4_K_M | 7.42 | 14.91 | 7.42 | 14.91 | 1 |
| Phi-4-mini-instruct.Q2_K | 7.21 | 11.16 | 7.21 | 11.16 | 1 |
| qwen2.5-3b-instruct-q5_k_m | 7.00 | 12.29 | 7.95 | 15.12 | 4 |
| gemma-2-2b-it-Q6_K | 6.92 | 16.88 | 10.27 | 21.37 | 6 |
| Llama-3.2-3B-Instruct-Q6_K | 5.38 | 12.03 | 7.33 | 15.45 | 6 |
| DeepSeek-R1-Distill-Qwen-1.5B-f16 | 5.35 | 7.19 | 5.35 | 7.19 | 1 |
| Gemmasutra-Mini-2B-v1-Q6_K | 5.19 | 10.59 | 5.19 | 10.59 | 1 |
| DeepSeek-R1-Distill-Qwen-1.5B-Fully-Uncensored.f16 | 4.83 | 6.89 | 5.07 | 7.02 | 2 |
| gemma-2-2b-it-IQ3_M | 4.55 | 7.82 | 4.55 | 7.82 | 1 |
| Phi-3.5-mini-instruct.Q4_K_M | 3.90 | 10.24 | 7.86 | 12.34 | 6 |
Head-to-Head Record
1–50 of 331 rows
1 / 7
Performance by App Version
ImprovedRegressed