Leaderboard
On-device LLM performance rankings powered by Glicko-2
Prism II
AndroidRank
#231
Rating
1,149
±19 RD
Win Rate
16.1%
Conservative Rating
1,112
TG Rating
1,141
PP Rating
1,229
Matches
784
Record
126W – 658L
Models Tested
| Model | TG Median (tok/s) | PP Median (tok/s) | TG Best | PP Best | Runs |
|---|---|---|---|---|---|
| llama-3.2-1b-instruct-q8_0 | 6.69 | 35.75 | 7.19 | 37.11 | 2 |
| LFM2-8B-A1B-Q4_K_M | 5.38 | 10.35 | 5.38 | 10.35 | 1 |
| granite-4.0-h-tiny-Q4_K_M | 5.38 | 13.49 | 5.38 | 13.49 | 1 |
| Qwen3-1.7B.Q3_K_M | 5.06 | 15.70 | 5.06 | 15.70 | 1 |
| Phi-3.5-mini-instruct.Q4_K_M | 2.93 | 8.92 | 2.93 | 8.92 | 1 |
| qwen2.5-1.5b-instruct-fp16 | 2.90 | 6.32 | 2.90 | 6.32 | 1 |
| qwen2.5-3b-instruct-q5_k_m | 2.84 | 7.39 | 2.84 | 7.39 | 1 |
| Llama-3.2-3B-Instruct-Q6_K | 2.19 | 6.77 | 2.19 | 6.77 | 1 |
| Mistral-7B-Instruct-v0.3.IQ4_XS | 1.95 | 4.38 | 1.95 | 4.38 | 1 |
| gemma-3-4b-it.Q4_K_M | 1.51 | 8.92 | 1.51 | 8.92 | 1 |
| Phi-4-mini-instruct.Q8_0 | 1.19 | 7.03 | 1.19 | 7.03 | 1 |
| google_gemma-3-1b-it-bf16 | 1.18 | 4.63 | 1.18 | 4.63 | 1 |
Head-to-Head Record
1–50 of 210 rows
1 / 5
Performance by App Version
ImprovedRegressed