Leaderboard

On-device LLM performance rankings powered by Glicko-2

POCO X7 Pro

Android

Rank

#68

Rating

1,736

±17 RD

Win Rate

72.8%

Conservative Rating

1,702

TG Rating

1,759

PP Rating

1,607

Matches

947

Record

689W – 258L

Models Tested

ModelTG Median (tok/s)PP Median (tok/s)TG BestPP BestRuns
Qwen3-0.6B-Q8_029.52139.5429.52139.541
DeepSeek-R1-Distill-Qwen-1.5B-Q4_K_M20.6036.7220.6036.721
SicariusSicariiStuff_Nano_Imp_1B-Q4_016.68115.4416.68115.441
SmolLM2-1.7B-Instruct-Q8_012.5050.3512.5050.351
llama-3.2-1b-instruct-q8_012.1457.0423.9377.602
unsloth.Q4_K_M11.0933.8811.0933.881
gemma-3-4B-it-QAT-Q4_010.2244.0010.2244.001
Phi-3.5-mini-instruct.Q4_K_M10.1514.0110.4316.554
Gemmasutra-Mini-2B-v1-Q3_K_M9.9717.489.9717.481
qwen2.5-3b-instruct-q5_k_m9.7313.4711.2217.465
Gemmasutra-Mini-2B-v1-Q6_K9.6822.5110.9229.034
gemma-2-2b-it-Q6_K8.3123.4510.0428.9213
gemma-3-4b-it-Q3_K_L7.0010.357.0010.351
Dolphin3.0-Llama3.1-8B-Q5_K_M4.876.254.876.251
gemma-3-1b-it.fp164.3711.244.3711.241
gemma-3n-E4B-it-Q5_K_M4.376.594.376.591
Phi-4-mini-instruct.Q8_04.2323.864.2323.861
Qwen3.5-4B-Uncensored-HauhauCS-Aggressive-Q4_K_M3.8724.903.8724.901
Qwen2.5-7B-Instruct.Q6_K3.755.423.755.421
Gemmasutra-Small-4B-v1a-Q8_03.6425.653.6425.651
DeepSeek-R1-Distill-Qwen-14B-IQ2_M1.491.921.491.921

Head-to-Head Record

Performance by App Version

ImprovedRegressed

Compare With