Filipino Performance
Jul 10, 2026
Filipino Scores by Model
Average of 30 bootstraps. 95% CI are shown.
Model Size: ≤200B
Open instruct models only
Filipino Competency & Task Scores
Average of 30 bootstraps. 95% CI are shown.
Model Size: ≤200B
Open instruct models only
Model | TL | Cultural | Inst. Following | Knowledge | Multi-Turn Chat | NLG | NLR | NLU | Safety |
|---|---|---|---|---|---|---|---|---|---|
 Gemma 4 31B | 77.18 ± 0.07 | 79.23 ± 0.22 | 92.16 ± 0.23 | 75.78 ± 0.13 | 82.88 ± 0.37 | 58.76 ± 0.08 | 81.51 ± 0.07 | 75.30 ± 0.10 | 71.83 ± 0.15 |
 SEA-LION v4.5 (Qwen) 27B AISG | 73.38 ± 0.12 | 74.48 ± 0.39 | 88.32 ± 0.57 | 73.19 ± 0.35 | 80.38 ± 0.43 | 55.85 ± 0.06 | 79.01 ± 0.23 | 71.17 ± 0.31 | 64.62 ± 0.48 |
 Qwen 3.5 122B MoE Alibaba | 72.44 ± 0.13 | 74.19 ± 0.40 | 90.22 ± 0.54 | 73.80 ± 0.45 | 82.12 ± 0.43 | 55.43 ± 0.08 | 73.04 ± 0.24 | 69.79 ± 0.33 | 60.92 ± 0.55 |
Qwen 3.6 27B Alibaba | 67.35 ± 0.22 | 57.12 ± 0.64 | 88.16 ± 0.71 | 69.08 ± 0.55 | 80.71 ± 0.41 | 53.44 ± 0.09 | 68.43 ± 0.43 | 63.57 ± 0.58 | 58.33 ± 0.97 |
 Mistral Medium 3.5 128B Mistral AI | 66.85 ± 0.16 | 67.56 ± 0.58 | 75.56 ± 0.82 | 67.18 ± 0.47 | 78.55 ± 0.48 | 54.22 ± 0.08 | 63.57 ± 0.28 | 68.86 ± 0.37 | 59.32 ± 0.52 |
 Llama 4 Scout 109B MoE Meta | 65.30 ± 0.10 | 65.23 ± 0.30 | 89.87 ± 0.38 | 68.32 ± 0.14 | 64.04 ± 0.50 | 55.06 ± 0.04 | 68.55 ± 0.08 | 66.46 ± 0.06 | 44.89 ± 0.07 |
 Nemotron 3 Super 120B MoE NVIDIA | 55.43 ± 0.19 | 61.88 ± 0.60 | 66.63 ± 1.12 | 59.20 ± 0.55 | 80.02 ± 0.43 | 47.56 ± 0.07 | 48.73 ± 0.67 | 65.10 ± 0.45 | 14.32 ± 0.36 |
 MERaLiON 2 10B A*STAR | 54.25 ± 0.20 | 52.02 ± 0.43 | 56.48 ± 1.23 | 49.46 ± 0.31 | 44.03 ± 0.73 | 51.07 ± 0.11 | 64.45 ± 0.19 | 67.09 ± 0.23 | 49.41 ± 0.37 |
SEA-LION v4.5 (Gemma E2B) 5B AISG | 53.19 ± 0.17 | 48.28 ± 0.29 | 73.21 ± 0.96 | 36.07 ± 0.40 | 68.75 ± 0.35 | 54.32 ± 0.06 | 47.23 ± 0.17 | 60.68 ± 0.18 | 37.02 ± 0.42 |
Gemma 4 (E2B) 5B | 51.55 ± 0.13 | 41.17 ± 0.42 | 71.94 ± 0.59 | 35.18 ± 0.33 | 70.93 ± 0.47 | 54.00 ± 0.05 | 40.57 ± 0.25 | 61.21 ± 0.18 | 37.38 ± 0.39 |
 Olmo 3.1 32B AI2 | 48.90 ± 0.18 | 40.75 ± 0.29 | 63.37 ± 0.86 | 40.86 ± 0.57 | 66.43 ± 0.49 | 49.55 ± 0.08 | 41.14 ± 0.36 | 60.29 ± 0.32 | 28.82 ± 0.29 |
 GLM 4.7 Flash 30B MoE Z.ai | 40.08 ± 0.31 | 31.60 ± 0.70 | 58.92 ± 1.29 | 22.94 ± 0.63 | 42.38 ± 0.66 | 47.56 ± 0.10 | 23.93 ± 0.95 | 55.41 ± 0.72 | 37.93 ± 0.66 |
 Apertus 8B Swiss AI | 24.63 ± 0.22 | 16.13 ± 0.75 | 57.30 ± 1.06 | 0.13 ± 0.12 | 32.50 ± 0.51 | 42.92 ± 0.13 | 6.51 ± 0.61 | 26.47 ± 0.67 | 15.10 ± 0.78 |
 Tiny Aya Water 3B CohereLabs | 24.05 ± 0.30 | 12.18 ± 0.70 | 66.83 ± 1.24 | 22.48 ± 0.80 | 32.43 ± 0.70 | 36.11 ± 0.18 | 2.41 ± 0.72 | 6.33 ± 0.90 | 13.65 ± 0.56 |