# Mixture of Experts
Latest news and articles about Mixture of Experts
Total: 3 articles found
NVIDIA’s Omni-Vision: Setting New Benchmarks for the Era of Autonomous AI Agents
NVIDIA has launched Nemotron 3 Nano Omni, a multimodal AI model utilizing a Mixture-of-Experts architecture to deliver 9x the efficiency of competing open models. Designed for autonomous agents, the model integrates text, video, and audio reasoning to enable real-time digital interaction and lower deployment costs.
Efficiency Over Scale: Bailing Unveils Ling-2.6-flash to Disrupt the Intelligence-Cost Curve
Bailing has launched Ling-2.6-flash, a 104B parameter model that uses Mixture of Experts (MoE) technology to activate only 7.4B parameters. It achieves benchmark parity with larger models while consuming only 10% of the tokens required by competitors like Nemotron-3-Super.
Alibaba’s Qianwen Open-Sources an 80B Coding Model Optimized for Agents and Local Development
Alibaba’s Qianwen has open‑sourced Qwen3‑Coder‑Next, an 80B parameter model designed for coding agents and local deployment that combines hybrid attention with MoE to lower inference costs. The release aims to accelerate enterprise adoption in China by enabling on‑premise use and customization, while raising questions about IP, safety and the infrastructure needed to realize claimed efficiency gains.