AI Model Achieves 99% Performance Using Just 12.5% of Experts
Breakthrough AI model EMO achieves 99% performance using just 12.5% of experts, enabling efficient deployment. Developed by Allen Institute & UC Berkeley.
-
-
NVIDIA Releases Nemotron 3 Nano Omni Multimodal AI Model
NVIDIA releases Nemotron 3 Nano Omni — a 30B open-weights multimodal AI model handling 100+ page docs and 5+ hrs of audio. Outperforms Alibaba’s Qwen3-Omni on key benchmarks. Available now on Hugging Face.
-
Google Releases Gemma 4 Open Source AI Models for On-Device Use
Google releases Gemma 4: four open source multimodal AI models (2.3B–31B params) supporting text, image, video & audio — all Apache 2.0, built for laptops, phones & edge devices.
