Leaderboard
On-device LLM performance rankings powered by Glicko-2
OnePlus 13
AndroidRank
#32
Rating
1,878
±14 RD
Win Rate
86.5%
Conservative Rating
1,849
TG Rating
1,902
PP Rating
1,806
Matches
1,314
Record
1136W – 178L
Models Tested
| Model | TG Median (tok/s) | PP Median (tok/s) | TG Best | PP Best | Runs |
|---|---|---|---|---|---|
| qwen2.5-coder-0.5b-instruct-q4_k_m | 75.93 | 137.14 | 75.93 | 137.14 | 1 |
| gemma-3-270m-it-F16 | 52.36 | 195.70 | 54.99 | 205.82 | 2 |
| gemma-3-1b-it.Q2_K | 47.73 | 143.54 | 47.73 | 143.54 | 1 |
| LFM2-1.2B-Q4_0 | 44.28 | 996.18 | 44.28 | 996.18 | 1 |
| DeepSeek-R1-Distill-Qwen-1.5B-IQ4_NL | 40.69 | 124.52 | 40.69 | 124.52 | 1 |
| Llama-3.2-1B-Instruct-Q4_0 | 33.34 | 287.70 | 33.34 | 287.70 | 1 |
| DeepSeek-R1-Distill-Qwen-1.5B-Q4_K_M | 33.06 | 59.14 | 33.06 | 59.14 | 1 |
| Yi-Coder-1.5B-Chat.Q4_K_M | 32.53 | 64.61 | 32.53 | 64.61 | 1 |
| llama-3.2-1b-instruct-q8_0 | 31.19 | 210.77 | 31.19 | 210.77 | 1 |
| Llama-3.2-1B-Instruct.IQ2_XS | 31.13 | 50.47 | 31.13 | 50.47 | 1 |
| Qwen2.5-1.5B-Instruct.Q8_0 | 29.67 | 126.35 | 29.67 | 126.35 | 1 |
| qwen2.5-1.5b-instruct-q8_0 | 26.14 | 153.92 | 26.14 | 153.92 | 1 |
| SmolLM2-1.7B-Instruct-Q8_0 | 24.08 | 109.42 | 24.08 | 109.42 | 1 |
| Llama-3.2-1B-Instruct.fp16 | 19.23 | 71.46 | 19.23 | 71.46 | 1 |
| gemma-2-2b-it-Q6_K | 14.84 | 30.15 | 16.18 | 66.26 | 7 |
| Phi-3-mini-128k-instruct.Q4_K_M | 14.78 | 21.23 | 14.78 | 21.23 | 1 |
| Phi-3.5-mini-instruct.Q4_K_M | 13.74 | 23.62 | 15.87 | 34.87 | 8 |
| Gemmasutra-Mini-2B-v1-Q6_K | 13.38 | 30.57 | 15.89 | 51.93 | 3 |
| Huihui-Qwen3.5-0.8B-abliterated.Q8_0 | 13.26 | 547.13 | 13.26 | 547.13 | 1 |
| Qwen3.5-2B.Q5_K_M | 12.73 | 70.05 | 12.73 | 70.05 | 1 |
| Llama-3.2-3B-Instruct-Q6_K | 12.16 | 24.81 | 13.48 | 54.64 | 8 |
| qwen2.5-3b-instruct-q5_k_m | 12.12 | 20.94 | 15.32 | 46.21 | 10 |
| Qwen3-4B-Instruct-2507-Q4_K_M | 11.58 | 24.15 | 11.58 | 24.15 | 1 |
| DeepSeek-R1-Distill-Qwen-7B-Q4_K_M | 9.82 | 12.91 | 9.82 | 12.91 | 1 |
| Reverb-7b.Q4_K_M | 9.09 | 14.38 | 9.09 | 14.38 | 1 |
| Dolphin3.0-Qwen2.5-3b-IQ2_M | 8.81 | 13.35 | 8.81 | 13.35 | 1 |
| Qwen3-8B.Q4_K_M | 8.47 | 12.35 | 8.47 | 12.35 | 1 |
| DeepSeek-R1-Distill-Llama-8B-Q4_K_M | 8.30 | 12.16 | 8.41 | 12.21 | 2 |
| copilot-7B-1-7M-unquantized.IQ4_XS | 7.85 | 11.85 | 7.85 | 11.85 | 1 |
| gemma-3-4b-it.Q8_0 | 7.18 | 45.98 | 7.18 | 45.98 | 1 |
| DeepSeek-R1-Distill-Qwen-7B-Q6_K | 6.71 | 10.00 | 6.71 | 10.00 | 1 |
| DeepSeek-R1-0528-Qwen3-8B-UD-Q6_K_XL | 6.16 | 12.96 | 6.16 | 12.96 | 1 |
| Qwen3-30B-A3B.Q3_K_M | 5.88 | 21.50 | 6.08 | 21.87 | 2 |
| Meta-Llama-3-8B-Instruct.IQ2_XS | 5.13 | 6.85 | 5.13 | 6.85 | 1 |
| open-hermes-2.5-neural-chat-3.1-frankenmerge-11b.Q8_0 | 4.09 | 11.62 | 4.09 | 11.62 | 1 |
| Qwen3.5-9B-Q8_0 | 3.36 | 17.09 | 3.36 | 17.09 | 1 |
| MiMo-7B-RL-Q4_K_M | 3.32 | 8.94 | 3.32 | 8.94 | 1 |
| DeepSeek-R1-Distill-Qwen-14B-Q4_K_M | 2.98 | 4.29 | 2.98 | 4.29 | 1 |
| Qwen3-14B.Q6_K | 2.81 | 5.78 | 2.81 | 5.78 | 1 |
| Qwen3-4B-Instruct-2507-F16 | 2.64 | 9.28 | 2.64 | 9.28 | 1 |
| DeepSeek-R1-Distill-Qwen-14B-Q8_0 | 1.04 | 8.40 | 1.04 | 8.40 | 1 |
Head-to-Head Record
1–50 of 332 rows
1 / 7
Performance by App Version
ImprovedRegressed