#Mixture-of-Experts

'Uni-MoE-2.0-Omni is an open omnimodal MoE on a Qwen2.5-7B core that unifies text, image, audio and video via a shared token interface, dynamic expert routing and staged training for robust multimodal understanding and generation.'

OpenAI's gpt-oss-20b plus NVIDIA RTX-optimized runtimes bring powerful, private LLMs to local PCs, enabling low-latency, customizable AI for students and organizations.

'Alibaba released Qwen3-Max, a 1T+ Mixture-of-Experts model with a production-ready thinking mode and notable day-one wins on coding and agentic benchmarks. Model Studio and Qwen Chat provide immediate access, with thinking runs requiring streaming output.'

'Baidu released ERNIE-4.5-21B-A3B-Thinking, a 21B-parameter MoE model with 3B active parameters and 128K context length designed for efficient multi-step reasoning.'

'Zhipu AI released GLM-4.5V, an open-source vision-language model that combines a 106B parameter MoE backbone with 12B active parameters, 64K token context and a tunable Thinking Mode for advanced multimodal reasoning.'

OpenAI introduces two powerful open-weight language models, gpt-oss-120B and gpt-oss-20B, allowing users to run advanced AI locally on laptops and phones with full customization and privacy.

'SmallThinker introduces a family of efficient large language models specifically designed for local device deployment, offering high performance with minimal memory and compute requirements. These models set new standards in on-device AI capabilities across multiple benchmarks and hardware constraints.'

Alibaba introduces Qwen3-MT, a next-generation multilingual machine translation model featuring cutting-edge architecture and reinforcement learning for high-quality, cost-efficient translations across 92+ languages.

Qwen has released Qwen3-Coder-480B-A35B-Instruct, the most powerful open agentic code model to date, featuring massive scale, extensive language support, and a new open-source CLI tool called Qwen Code.

DeepSeek-TNG introduces R1T2 Chimera, a new Assembly-of-Experts LLM that delivers twice the speed of R1-0528 and improved reasoning, available now under MIT license.

Baidu releases ERNIE 4.5, a series of open-source large language models scaling from 0.3 billion to 424 billion parameters, featuring advanced architectures and strong multilingual capabilities.

Mixture-of-Experts models selectively activate parts of AI networks, enabling huge parameter counts with efficient computation. This innovation is reshaping AI applications in language, vision, and recommendations.

IBM releases Granite 4.0 Tiny Preview, a compact open-source language model optimized for handling long-context and instruction-based tasks with impressive efficiency and performance.

Alibaba's Qwen3 introduces a new generation of large language models that excel in hybrid reasoning, multilingual understanding, and efficient scalability, setting new standards in AI performance.