The framework adds tools for testing and refining robotic AI systems, addressing a key bottleneck in making these models more practical.
LeRobot v0.6.0 is a software release focused on robot learning that introduces three main capabilities: world model policies that learn to predict future outcomes before acting, reward models that assess when a robot succeeds, and evaluation benchmarks to measure performance. The release addresses what the robotics field is asking: whether world models actually help robot policies work better. The update also adds new vision-language-action models, faster data loading with depth sensing support, automated language annotations for datasets, and deployment tools that can incorporate human corrections to improve training data.

OpenAI has released two new Realtime models in its API. They are named gpt-realtime-2.1 and gpt-realtime-2.1-mini. Both target low-latency voice and multimodal experiences. The mini model is the notable part of this release. It is a mini reasoning model for realtime voice. It ships at the same cost as the earlier gpt-realtime-mini. OpenAI also reduced p95 latency by at least 25% across Realtime voice models. That reduction comes from improved caching. What is GPT-Realtime-2.1-mini gpt-rea

Tencent’s Hy team released Hy3. Hy3 is a 295B-parameter Mixture-of-Experts (MoE) model. It activates only 21B parameters per token. The weights ship under the Apache License 2.0. Hy3 is aimed at reasoning, agentic workflows, and long-context tasks. What is Hy3? Hy3’s architecture contains a sparse MoE with 192 experts and top-8 routing. Only 8 experts fire per token, so compute stays low. The model also uses a Multi-Token Prediction (MTP) layer. MTP predicts several tokens
The open-source library's improvements could make AI models faster and more efficient to run on consumer hardware.
Want to go deeper than the news? Explore live, cohort-based AI courses taught by practitioners.
Browse AI courses on Maven