Leaderboard
On-device LLM performance rankings powered by Glicko-2
POCO F6
AndroidRank
#104
Rating
1,620
±15 RD
Win Rate
61.6%
Conservative Rating
1,590
TG Rating
1,627
PP Rating
1,575
Matches
1,195
Record
736W – 459L
Models Tested
| Model | TG Median (tok/s) | PP Median (tok/s) | TG Best | PP Best | Runs |
|---|---|---|---|---|---|
| gemma-3-270m-it-F16 | 20.19 | 99.42 | 20.19 | 99.42 | 1 |
| gemma-3-1b-it.Q8_0 | 15.62 | 73.54 | 15.62 | 73.54 | 1 |
| DeepSeek-R1-Distill-Qwen-1.5B-Q4_K_M | 13.94 | 61.81 | 13.94 | 61.81 | 1 |
| llama-3.2-1b-instruct-q8_0 | 13.05 | 61.91 | 14.70 | 94.33 | 4 |
| DeepSeek-R1-Distill-Qwen-1.5B-IQ2_M | 12.94 | 20.74 | 12.94 | 20.74 | 1 |
| Qwen3-1.7B.Q6_K | 12.02 | 27.67 | 12.02 | 27.67 | 1 |
| Parm2-Qwen2.5-3B.Q4_K_M | 9.30 | 24.50 | 9.30 | 24.50 | 1 |
| DeepSeek-R1-Distill-Qwen-1.5B-Q6_K | 9.06 | 32.13 | 12.60 | 32.41 | 2 |
| qwen2.5-1.5b-instruct-q8_0 | 9.05 | 36.36 | 9.05 | 36.36 | 1 |
| rombo-llm-v2.5-qwen-3b-q4_0 | 8.69 | 107.04 | 8.82 | 107.62 | 6 |
| qwen2.5-3b-instruct-q5_k_m | 8.32 | 15.17 | 8.32 | 15.17 | 1 |
| Phi-3.5-mini-instruct.Q4_K_M | 7.76 | 15.82 | 8.21 | 16.41 | 3 |
| gemma-2-2b-it-Q6_K | 6.65 | 16.35 | 7.06 | 17.55 | 2 |
| Qwen3-4B-Q4_K_M | 6.52 | 18.63 | 6.52 | 18.63 | 1 |
| Llama-3.2-3B-Instruct-Q6_K | 4.72 | 12.75 | 6.34 | 18.95 | 2 |
| Gemma-3-4B-VL-it-Gemini-Pro-Heretic-Uncensored-Thinking_Q6_k | 1.82 | 17.44 | 1.82 | 17.44 | 1 |
Head-to-Head Record
1–50 of 300 rows
1 / 6
Performance by App Version
ImprovedRegressed