Leaderboard

On-device LLM performance rankings powered by Glicko-2

iPhone 13 Pro Max

iOS

Rank

#24

Rating

1,901

±14 RD

Win Rate

88.6%

Conservative Rating

1,872

TG Rating

1,786

PP Rating

1,930

Matches

1,284

Record

1138W – 146L

Models Tested

ModelTG Median (tok/s)PP Median (tok/s)TG BestPP BestRuns
yi-ai-4b-chinese-it-v1-q6_k1359.4642173.881359.4642173.881
SmolLM2-135M-Instruct-Q8_0103.553206.11113.773399.063
google_gemma-3-270m-it-qat-Q8_063.45226.5463.45226.541
gemma-3-270m-it-F1653.89245.9656.06250.272
google_gemma-3-1b-it-qat-IQ3_M27.87574.1534.01670.992
DeepSeek-R1-Distill-Qwen-1.5B-Q4_027.61427.8027.61427.801
gemma-3-1b-it.Q5_K_M25.4042.2825.4042.281
gemma-3-1b-it.Q8_023.52748.2823.52748.281
llama-3.2-1b-instruct-q8_022.9954.2323.28638.083
DeepSeek-R1-Distill-Qwen-1.5B-IQ4_NL18.3626.8818.3626.881
qwen2.5-1.5b-instruct-q8_016.83235.2417.06433.742
DeepSeek-R1-Distill-Qwen-1.5B-Q4_K_M15.3030.1015.3030.101
gemma-3-1b-it.fp1613.64779.0713.64779.071
Qwen3.5-2B-IQ4_NL12.71241.9412.71241.941
DeepSeek-R1-Distill-Qwen-1.5B-IQ2_M12.3215.5112.3215.511
qwen2.5-3b-instruct-q5_k_m11.73165.9612.33172.985
Gemmasutra-Mini-2B-v1-Q6_K10.91187.0013.07232.522
Phi-3.5-mini-instruct.Q4_K_M9.72133.439.72133.431
Llama-3.2-3B-Instruct-Q6_K9.00140.969.48177.032
Llama-3.2-3B-Instruct-Q4_07.8113.957.8113.951
gemma-2-2b-it-Q6_K7.8013.6512.26200.415
Qwen3-4B-Instruct-2507.Q2_K6.998.896.998.891
Dolphin3.0-Llama3.2-3B-Q6_K6.67120.256.67120.251
Qwen3-4B.Q3_K_L5.907.315.907.311
gemma-3-4b-it.Q6_K3.0539.793.0539.791

Head-to-Head Record

Performance by App Version

ImprovedRegressed

Compare With