Leaderboard

On-device LLM performance rankings powered by Glicko-2

Galaxy S24+

Android

Rank

#70

Rating

1,724

±14 RD

Win Rate

71.6%

Conservative Rating

1,695

TG Rating

1,738

PP Rating

1,729

Matches

1,301

Record

931W – 370L

Models Tested

ModelTG Median (tok/s)PP Median (tok/s)TG BestPP BestRuns
tinygemma3-Q8_0237.711134.41237.711134.411
gemma-3-1b-it-q8_028.82147.7728.82147.771
llama-3.2-1b-instruct-q8_020.72113.4424.77123.115
ERNIE-4.5-0.3B-PT-F1620.3062.5120.3062.511
gemma-3-1b-it.Q6_K16.53101.9016.53101.901
Qwen3-1.7B-Q8_012.5360.7512.5360.751
DeepSeek-R1-Distill-Qwen-1.5B-Q8_012.3273.6017.5178.502
qwen2.5-3b-instruct-q4_k_m11.7947.8411.7947.841
DeepSeek-R1-Distill-Qwen-1.5B-Q4_K_M9.9847.659.9847.651
gemma-3-4b-it.Q4_K_M9.5123.359.5123.351
Qwen3-4B.Q4_K_M9.3320.939.3320.931
Gemmasutra-Mini-2B-v1-Q6_K9.1626.439.1626.431
qwen2.5-3b-instruct-q5_k_m8.5816.109.5920.585
DeepSeek-R1-Distill-Qwen-1.5B-IQ2_M8.5425.568.5425.561
gemma-2-2b-it-Q6_K8.5326.449.2037.705
Phi-3.5-mini-instruct.Q4_K_M8.2322.2310.8627.733
Llama-3.2-3B-Instruct-Q6_K8.0819.4411.6123.124
gemma-3-4b-it.Q8_07.6927.447.6927.441
medgemma-4b-it-IQ4_NL6.9841.346.9841.341
Gemma-3-4B-VL-it-Gemini-Pro-Heretic-Uncensored-Thinking_Q8_06.8544.856.8544.851
DeepSeek-R1-Distill-Qwen-7B-Q4_K_M6.3712.286.3712.281
Phi-4-mini-instruct.Q8_05.7322.205.7322.201
Qwen3-4B-Instruct-2507-Q8_05.3619.485.3619.481
Janus-Pro-7B-LM.Q2_K5.187.395.187.391
SmolLM2-1.7B-Instruct-Q8_04.5832.854.8433.062
DeepSeek-R1-Distill-Llama-8B-Abliterated.Q5_K_S3.635.753.635.751
Qwen3.5-2B-Uncensored-HauhauCS-Aggressive-BF163.038.903.038.901
Qwen3.5-9B-Q4_02.566.572.566.571
LFM2.5-1.2B-Instruct-Q4_K_M1.2643.021.2643.021

Head-to-Head Record

Performance by App Version

ImprovedRegressed

Compare With