Leaderboard

On-device LLM performance rankings powered by Glicko-2

Xiaomi 14

Android

Rank

#46

Rating

1,805

±16 RD

Win Rate

79.4%

Conservative Rating

1,773

TG Rating

1,830

PP Rating

1,744

Matches

1,053

Record

836W – 217L

Models Tested

ModelTG Median (tok/s)PP Median (tok/s)TG BestPP BestRuns
Qwen2-0.5B-Instruct-Q8_050.46322.0250.46322.021
tinyllama-1.1b-chat-v1.0.Q2_K34.6050.8434.6050.841
hunyuan-0.5b-instruct-q8_032.42148.8632.42148.861
llama-3.2-1b-instruct-q8_025.01133.0825.19133.202
qwen2.5-1.5b-instruct-q8_019.4778.7919.4778.791
DeepSeek-R1-Distill-Qwen-1.5B-IQ4_NL18.5376.4518.5376.451
DeepSeek-R1-Distill-Qwen-1.5B-IQ2_M17.5127.1618.0827.832
Phi-4-mini-instruct-Q4_K_M13.9223.1213.9223.121
gemma-3-4B-it-QAT-Q4_012.9663.8812.9663.881
Phi-3.5-mini-instruct.Q4_K_M12.7121.5612.7121.561
gemma-2-2b-it-Q6_K12.5226.9612.5226.961
gemma-3-4b-it-Q4_K_M12.1326.0312.1326.031
qwen2.5-3b-instruct-q5_k_m11.9819.6012.3119.842
Llama-3.2-3B-Instruct-Q6_K11.3520.3711.9822.045
Llama-3.2-3B-Instruct-uncensored-Q8_011.1436.3811.1436.381
DeepSeek-R1-Distill-Qwen-1.5B-Q5_K_M9.3821.669.3821.661
DeepSeek-R1-Distill-Qwen-1.5B-UD-IQ4_XS8.0223.9411.1325.972
BgGPT-Gemma-2-2B-IT-v1.0.Q4_K_S7.4420.247.4420.241
DeepSeek-R1-Distill-Qwen-7B-Q4_K_M6.4311.146.4311.141
Qwen2.5-VL-3B-Instruct-bf16-q4_k5.8819.605.8819.601
MiMo-7B-RL-Q4_K_M5.719.895.719.891
Qwen2.5-7B-Instruct.Q5_K_M4.836.684.836.681
Qwen3.5-0.8B-Q4_K_M3.70115.333.70115.331
Meta-Llama-3.1-8B-Instruct-IQ3_XS3.224.153.224.151
DeepSeek-R1-Distill-Qwen-7B-Q6_K2.195.992.195.991

Head-to-Head Record

Performance by App Version

ImprovedRegressed

Compare With