Researchers have developed a lightweight memory module for AI agents that retains information across long interactions without expanding the context window or relying on external retrieval systems. Called delta-mem, the system adds just 0.12% of additional parameters to a base language model while outperforming much larger alternatives on memory-intensive tasks, Ben Dickson reports for VentureBeat. …
Alibaba has released Qwen3.7-Max, a proprietary AI model built specifically for autonomous, long-running tasks. The model is available exclusively through Alibaba Cloud’s paid API and is not open source, marking a significant shift from the company’s previous approach. The Qwen Team, Alibaba’s AI research group, says the model completed a continuous 35-hour engineering task entirely …
Google has announced Gemini Spark, a personal AI agent that runs continuously in the cloud and completes tasks without requiring the user to be present. The company unveiled the product at its annual I/O developer conference. Spark is powered by Gemini 3.5 Flash and connects directly to Google Workspace apps including Gmail, Docs, and Slides. …
Anthropic has launched Claude for Small Business, a package of software integrations and ready-made workflows supposed to bring its AI assistant into the daily operations of small and mid-sized companies. The product runs inside Claude Cowork, Anthropic’s task-automation platform, and connects to tools many small businesses might already use. The launch includes integrations with Intuit …
Google is developing a new AI agent for its Gemini platform that can perform tasks independently on behalf of users. Hilary Langley reports for Business Insider that the agent, internally named “Remy,” is currently being tested by Google employees. According to an internal document reviewed by Business Insider, Remy runs within a staff-only version of …
Mistral has launched remote coding agents and a new AI model called Mistral Medium 3.5, marking a significant step toward autonomous software development. Mistral explains that the system allows coding tasks to run in the cloud without requiring a developer to stay at their computer. The new model is a dense 128-billion-parameter system with a …
Nvidia has launched Nemotron 3 Nano Omni, an open AI model that combines text, vision and audio processing in a single system. Most existing AI agent systems rely on separate models for each modality, which increases latency and cost. Nvidia says its new model eliminates that fragmentation. The model uses a hybrid mixture-of-experts architecture with …
Xiaomi has launched two new open source artificial intelligence models: MiMo-V2.5 and MiMo-V2.5-Pro. Carl Franzen reports for VentureBeat that both models are available under the permissive MIT License, meaning businesses and developers can use, modify, and deploy them commercially without restrictions. The models are designed for so-called agentic tasks, where an AI system autonomously completes …
China’s top economic planning body, the National Development and Reform Commission (NDRC), has ordered Meta to reverse its acquisition of AI startup Manus. The deal was valued at over $2 billion when Meta completed it in late 2025. The NDRC said it would “prohibit foreign investment in Manus in accordance with laws and regulations” and …
Microsoft has expanded its artificial intelligence assistant Copilot with new capabilities that allow it to take direct, multi-step actions inside Word, Excel, and PowerPoint. Sumit Chauhan, President of the Office Product Group, writes that these so-called agentic features are now generally available and the default experience for subscribers. Previously, Copilot could answer questions but struggled …