Leaderboard
On-device LLM performance rankings powered by Glicko-2
Galaxy S24 FE
AndroidRank
#85
Rating
1,696
±16 RD
Win Rate
68.9%
Conservative Rating
1,664
TG Rating
1,696
PP Rating
1,722
Matches
1,046
Record
721W – 325L
Models Tested
| Model | TG Median (tok/s) | PP Median (tok/s) | TG Best | PP Best | Runs |
|---|---|---|---|---|---|
| LFM2.5-1.2B-Instruct-Q4_K_M | 33.65 | 130.84 | 33.65 | 130.84 | 1 |
| Qwen3-0.6B-Q8_0 | 23.43 | 118.49 | 23.43 | 118.49 | 1 |
| llama-3.2-1b-instruct-q8_0 | 15.18 | 119.44 | 15.18 | 119.44 | 1 |
| LFM2.5-1.2B-Thinking-Q8_0 | 13.28 | 107.34 | 13.28 | 107.34 | 1 |
| SmolLM2-1.7B-Instruct-Q8_0 | 13.10 | 66.53 | 13.42 | 69.30 | 2 |
| Qwen3-1.7B-Q8_0 | 12.26 | 53.20 | 12.80 | 53.34 | 2 |
| Qwen3.5-0.8B-UD-Q8_K_XL | 9.83 | 71.47 | 9.83 | 71.47 | 1 |
| gemma-2-2b-it-q6_k | 8.98 | 21.48 | 8.98 | 21.48 | 1 |
| DeepSeek-R1-Distill-Qwen-1.5B-IQ2_M | 8.45 | 20.78 | 8.45 | 20.78 | 1 |
| Phi-3.5-mini-instruct.Q4_K_M | 8.37 | 16.70 | 8.71 | 33.84 | 4 |
| qwen2.5-3b-instruct-q5_k_m | 8.24 | 17.66 | 8.24 | 17.66 | 1 |
| gemma-2-2b-it-Q6_K | 7.75 | 24.28 | 7.97 | 27.99 | 2 |
| Llama-3.2-3B-Instruct-Q6_K | 7.73 | 25.35 | 7.73 | 25.35 | 1 |
Head-to-Head Record
1–50 of 285 rows
1 / 6
Performance by App Version
ImprovedRegressed