Leaderboard
On-device LLM performance rankings powered by Glicko-2
Galaxy Note10+
AndroidRank
#180
Rating
1,357
±20 RD
Win Rate
36.2%
Conservative Rating
1,316
TG Rating
1,368
PP Rating
1,271
Matches
650
Record
235W – 415L
Models Tested
| Model | TG Median (tok/s) | PP Median (tok/s) | TG Best | PP Best | Runs |
|---|---|---|---|---|---|
| gemma-3-1b-it-q4_0 | 10.55 | 55.00 | 10.55 | 55.00 | 1 |
| gemma-3-1b-it-Q8_0 | 10.33 | 44.26 | 10.33 | 44.26 | 1 |
| llama-3.2-1b-instruct-q8_0 | 9.90 | 31.97 | 10.71 | 39.18 | 8 |
| gemma-3-1b-it-f16 | 6.09 | 12.25 | 6.09 | 12.25 | 1 |
| qwen2.5-3b-instruct-q8_0 | 4.17 | 11.36 | 4.17 | 11.36 | 1 |
| Llama-3.2-3B-Instruct-Q8_0 | 4.12 | 11.76 | 4.12 | 11.76 | 1 |
| gemma-2-2b-it-Q6_K | 3.98 | 11.61 | 4.62 | 22.15 | 3 |
| Gemma-3-it-4B-Uncensored-D_AU-IQ4_NL | 3.83 | 11.60 | 3.83 | 11.60 | 1 |
| gemma-3-4b-it-q4_0 | 3.58 | 14.80 | 3.58 | 14.80 | 1 |
| gemma-3-4b-it-Q8_0 | 3.43 | 10.77 | 3.43 | 10.77 | 1 |
| Llama-3.2-3B-Instruct-Q6_K | 2.01 | 5.29 | 2.04 | 5.37 | 2 |
| Meta-Llama-3-8B-Instruct.Q8_0 | 1.63 | 3.65 | 1.63 | 3.65 | 1 |
| qwen2.5-3b-instruct-q5_k_m | 1.28 | 4.90 | 2.48 | 7.19 | 2 |
| saiga_nemo_12b.Q2_K | 1.19 | 2.66 | 1.19 | 2.66 | 1 |
| gemma-3-12b-it-q4_0 | 1.06 | 4.60 | 1.06 | 4.60 | 1 |
| GigaChat-20B-A3B-instruct-v1.5.Q2_K | 0.99 | 4.70 | 0.99 | 4.70 | 1 |
| saiga_gemma3_12b.Q4_K_M | 0.69 | 2.95 | 0.69 | 2.95 | 1 |
| saiga_yandexgpt_8b.Q6_K | 0.57 | 3.02 | 0.57 | 3.02 | 1 |
Head-to-Head Record
1–50 of 214 rows
1 / 5
Performance by App Version
ImprovedRegressed