ilmu 0.1 is a Malaysian LLM intelligence model, built by Malaysians for Malaysians.
ilmu 0.1 outperforms other models globally on the Malay MMLU benchmark.
| Model | Language | Humanities | STEM | Social Sciences | Others | Average Accuracy |
|---|---|---|---|---|---|---|
| ilmu 0.1 (YTL) | 87.77 | 89.26 | 86.66 | 85.27 | 86.4 | 86.98 |
| GPT-4o (OpenAI) | 87.12 | 88.12 | 83.83 | 82.58 | 83.09 | 84.98 |
| Llama-3.1-70B (Meta) | 78.75 | 82.59 | 78.96 | 77.2 | 75.32 | 78.44 |
| SahabatAI (IndoSat + Goto) | 77.23 | 75.24 | 76.22 | 70.9 | 71.82 | 74.03 |
| Sea-lionv3 (AI-SG) | 75.56 | 73.27 | 76.01 | 69.85 | 71.26 | 72.82 |
| Sailor-14B-Chat (SEA AI Lab + SUDT) | 78.4 | 72.88 | 69.63 | 69.47 | 68.67 | 72.29 |
| Llama-3.3-70B (Meta) | 78.82 | 80.46 | 78.71 | 75.79 | 73.85 | 77.38 |
| Mallam 2.5 Small (Mesolitica) | 73 | 71 | 70 | 72 | 70 | 71.53 |
| SeaLLMs-v3-7B-Chat (Alibaba Group) | 67.67 | 64.69 | 69.18 | 62.43 | 63.68 | 65.1 |
| Merdeka-LLM (Agmo) | 56.92 | 60.36 | 57.63 | 56.82 | 55.1 | 57.28 |
ilmu 0.1 successfully passes both Malaysian PT3 and SPM national exams in the Malay language (Bahasa Melayu) with As, enabling it to support all industries including finance, education, healthcare, and public services, effectively in localisation.
We bring an inquisitive mindset to every problem, and transform them into innovative opportunities with infinite potential. Get in touch with us to experience Malaysian intelligence.