Vietnamese Performance
Vietnamese Scores by Model
Average of 2000 bootstraps. 95% CI are shown.
Model Size: ≤200B
Open instruct models only
Vietnamese Competency & Task Scores
Average of 2000 bootstraps. 95% CI are shown.
Model Size: ≤200B
Open instruct models only
Model | VI | Cultural | Inst. Following | Knowledge | Multi-Turn Chat | NLG | NLR | NLU | Safety |
|---|---|---|---|---|---|---|---|---|---|
 Gemma 4 31B | 77.09 [-1.71, +1.76] | 83.60 [-9.72, +9.23] | 95.34 [-3.91, +3.23] | 75.50 [-4.84, +4.75] | 85.63 [-3.54, +3.31] | 55.62 [-0.89, +0.87] | 78.89 [-2.54, +2.40] | 69.96 [-4.29, +4.54] | 72.18 [-4.25, +4.03] |
 SEA-LION v4.5 (Qwen) 27B AISG | 74.52 [-1.53, +1.63] | 80.64 [-8.46, +8.27] | 92.65 [-4.67, +3.90] | 75.68 [-4.60, +4.53] | 86.11 [-3.52, +3.63] | 54.34 [-0.79, +0.79] | 75.16 [-2.34, +2.35] | 66.34 [-4.35, +4.37] | 65.21 [-3.91, +3.79] |
 Qwen 3.5 122B MoE Alibaba | 74.33 [-1.54, +1.54] | 81.60 [-7.55, +7.56] | 90.47 [-5.35, +4.65] | 76.26 [-4.72, +4.57] | 86.58 [-3.23, +3.32] | 54.95 [-0.79, +0.77] | 71.39 [-2.49, +2.33] | 67.62 [-4.22, +4.27] | 65.77 [-3.64, +3.47] |
Qwen 3.6 27B Alibaba | 72.00 [-1.51, +1.55] | 77.46 [-7.71, +7.63] | 87.04 [-5.38, +5.10] | 75.25 [-4.25, +4.26] | 86.87 [-3.03, +3.05] | 54.17 [-0.74, +0.71] | 71.49 [-2.31, +2.29] | 65.79 [-4.03, +4.04] | 57.91 [-3.71, +3.77] |
 Mistral Medium 3.5 128B Mistral AI | 66.33 [-1.86, +1.78] | 48.13 [-10.09, +9.75] | 85.48 [-5.60, +5.11] | 68.81 [-5.27, +5.27] | 81.14 [-4.28, +4.15] | 54.12 [-0.73, +0.73] | 65.02 [-2.48, +2.38] | 68.67 [-4.31, +4.25] | 59.25 [-4.73, +4.55] |
 Llama 4 Scout 109B MoE Meta | 63.92 [-1.95, +2.02] | 53.95 [-10.89, +11.05] | 90.45 [-4.98, +4.43] | 64.34 [-5.67, +5.70] | 68.77 [-4.80, +4.61] | 54.61 [-0.84, +0.78] | 58.27 [-2.81, +2.84] | 67.45 [-4.61, +4.57] | 53.51 [-5.72, +5.69] |
 Nemotron 3 Super 120B MoE NVIDIA | 61.93 [-1.80, +1.65] | 58.31 [-8.09, +7.72] | 78.51 [-6.60, +6.02] | 62.56 [-4.69, +4.56] | 84.03 [-3.22, +3.18] | 50.64 [-0.66, +0.64] | 59.30 [-2.57, +2.35] | 60.70 [-3.84, +3.66] | 41.37 [-4.26, +4.37] |
SEA-LION v4.5 (Gemma E2B) 5B MoE AISG | 53.31 [-2.04, +2.01] | 31.87 [-9.12, +9.24] | 85.44 [-5.92, +5.51] | 38.97 [-5.97, +5.87] | 70.09 [-5.52, +5.27] | 53.00 [-0.73, +0.71] | 46.29 [-3.86, +3.72] | 53.58 [-4.56, +4.52] | 47.19 [-5.92, +5.81] |
 MERaLiON 2 10B A*STAR | 51.89 [-1.81, +1.85] | 50.65 [-7.78, +8.30] | 63.93 [-6.90, +6.43] | 27.41 [-5.24, +5.43] | 50.51 [-4.86, +4.97] | 52.58 [-0.73, +0.71] | 56.05 [-3.12, +2.91] | 55.45 [-4.04, +3.85] | 58.56 [-5.29, +4.88] |
 Olmo 3.1 32B AI2 | 51.07 [-1.89, +2.01] | 33.95 [-9.34, +9.17] | 81.14 [-6.73, +5.89] | 31.23 [-5.94, +5.81] | 69.95 [-4.87, +4.84] | 51.36 [-0.77, +0.76] | 41.53 [-3.72, +3.58] | 64.92 [-4.02, +4.23] | 34.47 [-4.80, +5.26] |
Gemma 4 (E2B) 5B | 49.85 [-1.85, +1.86] | 13.31 [-6.53, +7.38] | 83.15 [-6.61, +5.90] | 36.22 [-6.06, +6.09] | 71.96 [-5.40, +5.31] | 52.55 [-0.69, +0.66] | 43.99 [-3.96, +3.59] | 49.43 [-4.37, +4.27] | 48.20 [-5.88, +5.93] |
 GLM 4.7 Flash 30B MoE Z.ai | 44.11 [-1.38, +1.44] | 15.07 [-3.42, +3.61] | 71.30 [-5.94, +5.73] | 34.57 [-3.99, +4.30] | 60.77 [-5.10, +4.98] | 51.20 [-0.71, +0.68] | 18.38 [-1.56, +1.72] | 59.30 [-3.83, +3.67] | 42.31 [-4.13, +4.22] |
 Apertus 8B Swiss AI | 34.46 [-1.37, +1.43] | 0.00 [-0.00, +0.00] | 68.71 [-6.33, +6.18] | 28.99 [-4.82, +4.89] | 42.41 [-5.77, +6.12] | 48.42 [-0.75, +0.70] | 23.90 [-2.14, +1.99] | 43.81 [-3.58, +3.47] | 19.47 [-3.22, +3.70] |
 Tiny Aya Water 3B CohereLabs | 32.15 [-1.36, +1.29] | 0.00 [-0.00, +0.00] | 65.89 [-6.49, +6.37] | 30.80 [-5.14, +5.07] | 35.17 [-4.74, +4.75] | 39.23 [-0.63, +0.58] | 20.62 [-2.33, +2.24] | 36.76 [-3.11, +3.03] | 28.76 [-3.51, +3.71] |