language model

AI 2 min read

MiniMax to Open-Source Groundbreaking 2.7 Trillion Parameter AI Model This Year

Chinese AI startup MiniMax is set to release a monumental 2.7 trillion parameter language model as open source, potentially reshaping the AI landscape.

The Decoder

30 Jul 2026, 09:46 GST

AI 2 min read

Tencent's Hy3 Model Promises High Performance with Fewer Resources

Tencent's new open-source language model, Hy3, boasts 295 billion parameters but activates only 21 billion at a time, claiming performance on par with much larger models.

The Decoder · 30 Jul 2026 2 min read