Leaderboard
On-device LLM performance rankings powered by Glicko-2
Galaxy S22
AndroidRank
#122
Rating
1,539
±16 RD
Win Rate
53.8%
Conservative Rating
1,507
TG Rating
1,541
PP Rating
1,410
Matches
1,026
Record
552W – 474L
Models Tested
| Model | TG Median (tok/s) | PP Median (tok/s) | TG Best | PP Best | Runs |
|---|---|---|---|---|---|
| Llama-3.2-1B-Instruct-Q8_0 | 16.65 | 44.30 | 16.65 | 44.30 | 1 |
| smollm2-360m-instruct-q8_0 | 15.18 | 87.07 | 15.18 | 87.07 | 1 |
| llama-3.2-1b-instruct-q8_0 | 12.03 | 54.10 | 15.99 | 58.70 | 2 |
| gemma-3-1b-it.Q4_K_M | 10.44 | 41.64 | 10.79 | 41.91 | 3 |
| SmolLM2-1.7B-Instruct-Q8_0 | 6.91 | 24.01 | 6.91 | 24.01 | 1 |
| Hermes-3-Llama-3.2-3B.Q4_K_M | 6.80 | 13.15 | 6.80 | 13.15 | 1 |
| google_gemma-3n-E2B-it-IQ4_XS | 6.62 | 13.39 | 6.62 | 13.39 | 1 |
| qwen2.5-3b-instruct-q5_k_m | 6.45 | 11.31 | 6.45 | 11.31 | 1 |
| Phi-3.5-mini-instruct.Q4_K_M | 5.80 | 10.53 | 6.18 | 10.80 | 4 |
| qwen2.5-1.5b-instruct-q8_0 | 5.26 | 44.52 | 8.74 | 60.35 | 2 |
| Qwen3.5-0.8B-UD-Q8_K_XL | 4.70 | 38.21 | 4.70 | 38.21 | 1 |
| gemma-2-2b-it-Q6_K | 4.52 | 11.12 | 5.03 | 12.37 | 4 |
Head-to-Head Record
1–50 of 264 rows
1 / 6
Performance by App Version
ImprovedRegressed