Leaderboard
On-device LLM performance rankings powered by Glicko-2
Galaxy S23+
AndroidRank
#162
Rating
1,433
±22 RD
Win Rate
43.5%
Conservative Rating
1,389
TG Rating
1,442
PP Rating
1,487
Matches
554
Record
241W – 313L
Models Tested
| Model | TG Median (tok/s) | PP Median (tok/s) | TG Best | PP Best | Runs |
|---|---|---|---|---|---|
| chatgpt-5-q8_0 | 38.12 | 304.74 | 38.12 | 304.74 | 1 |
| qwen2.5-0.5b-instruct-q8_0 | 27.84 | 156.74 | 27.84 | 156.74 | 1 |
| gemma-3-1B-it-QAT-Q4_0 | 22.41 | 102.48 | 22.41 | 102.48 | 1 |
| Qwen3-0.6B-Q5_K_M | 20.43 | 53.45 | 20.43 | 53.45 | 1 |
| gemma-3-1b-it.Q4_K_M | 15.77 | 55.97 | 15.77 | 55.97 | 1 |
| Qwen2.5-1.5B-Instruct-Q4_K_M | 15.64 | 35.38 | 15.64 | 35.38 | 1 |
| Gemmasutra-Mini-2B-v1-Q6_K | 6.33 | 16.30 | 6.33 | 16.30 | 1 |
| gemma-2-2b-it-Q6_K | 5.68 | 16.95 | 7.90 | 17.08 | 2 |
| Llama-3.2-3B-Instruct-Q6_K | 2.04 | 11.32 | 2.04 | 11.32 | 1 |
| SmolLM2-1.7B-Instruct-Q8_0 | 0.46 | 7.12 | 0.46 | 7.12 | 1 |
Head-to-Head Record
1–50 of 160 rows
1 / 4
Performance by App Version
ImprovedRegressed