{
  "id": 351589,
  "title": "AMD เปิดตัว Instella-MoE-16B-A3B — โมเดล AI ที่เทรนด้วย GPU ของตัวเอง ไม่พึ่ง Nvidia",
  "url": "https://urgent.news/2026/08/09/amd-instella-moe-16b-a3b-ai-gpu-nvidia",
  "topic": "ai",
  "section": "AI",
  "published": "2026-08-09T06:33:23.000Z",
  "source": {
    "name": "Dev.to",
    "slug": "dev-to",
    "url": "https://dev.to/sarantoon/amd-epidtaw-instella-moe-16b-a3b-omedl-ai-thiiethrndwy-gpu-khngtaweng-aimphueng-nvidia-4868"
  },
  "original_language": "th",
  "account": "AMD has unveiled Instella-MoE-16B-A3B, an AI model that harnesses AMD's own GPUs without relying on Nvidia. This development comes from Nokka (Nok-ga). Unlike many AI models, Instella-MoE-16B-A3B is not just free; it demonstrates that AMD GPUs can train top-tier AI models without Nvidia's GPUs. The model, which has 16 billion parameters, employs a technique called Mixture-of-Experts (MoE) that activates only 2.8 billion specialized experts at a time, significantly speeding up computation. This approach, akin to a company assigning tasks to specialized departments, enhances performance while maintaining broad knowledge. AMD trained this model exclusively on its Instinct MI300X and MI325X GPUs, avoiding Nvidia's CUDA. This achievement marks a significant shift, as it shows AMD's GPUs can power cutting-edge AI models independently. AMD has also made the model's weights and training code freely available, using licenses that allow research and commercial use under certain conditions. This move opens up AI model development beyond the confines of Nvidia, promoting competition and innovation in the GPU ecosystem.",
  "summary": "AMD เปิดตัว Instella-MoE-16B-A3B — โมเดล AI ที่เทรนด้วย GPU ของตัวเอง ไม่พึ่ง Nvidia โดย Nokka (นก-กา) | 5 สิงหาคม 2569 AMD เพิ่งปล่อยโมเดล AI ตัวใหม่ที่สร้างความฮือฮาในวงการ เพราะไม่ได้เป็นเพียง \"โมเดลฟรีอีกตัว\" แต่มันคือการพิสูจน์ว่า GPU ของ AMD ก็เทรน AI ระดับท็อปได้จริง โดยไม่ต้องพึ่ง Nvidia [1][2] โมเดลชื่อ Instella-MoE-16B-A3B มีขนาด 16 พันล้านพารามิเตอร์…",
  "key_points": [
    "AMD unveils Instella-MoE-16B-A3B AI model using AMD GPUs",
    "Model employs Mixture-of-Experts technique for efficiency",
    "AMD trained model exclusively on Instinct MI300X and MI325X GPUs"
  ],
  "editors_take": "AMD's release of Instella-MoE-16B-A3B shows its GPUs can train top-tier AI models independently, shifting the landscape by reducing reliance on Nvidia and opening up AI development to more competition.",
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}