Each week, Quartz rounds up product launches, updates, and funding news from artificial intelligence-focused startups and companies.
Here’s what’s going on this week in the ever-evolving AI industry.
Google (GOOGL) announced Gemini 2.5 this week, which it called its “most intelligent AI model.” Gemini 2.5 Pro Experimental is the first release of Gemini 2.5, and is Google’s “most advanced model for complex tasks.” The Gemini 2.5 models are “thinking models” that can reason through inquiries before responding.
Following its first “thinking” model, Gemini 2.0 Flash Thinking, Gemini 2.5 combines an “enhanced base model with improved post-training,” which includes techniques such as reinforcement learning and chain-of-thought prompting.
Gemini 2.5 Pro Experimental is at the top of the crowdsourced Chatbot Arena rankings, and demonstrates strong reasoning and coding capabilities, Google said. The model is available to Gemini Advanced users in the app.
Alibaba (BABA) Cloud launched its Qwen2.5-Omni-7B AI model this week. The multimodal AI model can process text, images, audio, and videos, and can generate text and natural speech responses. The compact model “makes it the perfect foundation for developing agile, cost-effective AI agents,” Alibaba said.
The open-source model is available through Hugging Face and GitHub.
DeepSeek released an upgraded version of its V3 large language model this week. The Chinese AI startup said the new version delivers a “major boost in reasoning performance, stronger front-end development skills, and smarter tool-use capabilities.”
DeepSeek-V3-0324 is available on Hugging Face.
Agentic AI company Landbase announced its new Applied AI Lab this week, which will be based in Silicon Valley. The team, which includes talent from places such as Stanford, Meta, and NASA, will focus on agentic AI.
The Applied AI Lab try accelerating “workflow automation, data intelligence and reinforcement learning” innovations.
“Innovation is only as great as the team behind you,” Daniel Saks, CEO of Landbase, said in a statement. “That’s why we created the Applied AI Lab and brought on world-class AI and ML experts to continue advancing and refining the GTM-1 Omni model.”
The GTM-1 Omni model is the first and only agentic AI model that was built for go-to-market purposes, according to Landbase.
The lab will focus on models for planning and decision-making, models for generating messaging, and prediction and reward models.
AI solutions company webAI and enterprise cloud provider MacStadium announced a partnership this week on a new platform to deploy large AI models with silicon technology from Apple (AAPL).
The collaboration will enable more computational efficiency and performance for businesses facing hardware constraints.
“This collaboration with webAI represents a significant milestone in the evolution of AI for the enterprise,” Ken Tacelli, CEO at MacStadium, said in a statement. “By combining our macOS cloud expertise with webAI’s interconnected model approach, we’re creating a platform that will fundamentally change how organizations develop and deploy advanced AI systems on Apple silicon.”
The post Google’s Gemini 2.5, Alibaba’s new Qwen, and upgraded DeepSeek V3: This week’s AI launches appeared first on Quartz.
