SCRFrench startup Mistral AI has announced the release of Mistral Small 3, a 24-billion-parameter language model that the company claims matches the performance of models three times its size. According to Mistral AI, the new model achieves 81% accuracy on standard benchmarks while processing 150 tokens per second, making it comparable to Meta’s Llama 3.3 …
Alibaba’s Qwen team has launched two significant AI models – Qwen2.5-VL and Qwen2.5-Max – that demonstrate advanced capabilities in various tasks. According to the company, these models can perform text and image analysis, control computers and mobile devices, and compete with established AI systems from OpenAI, Anthropic, and Google on multiple benchmarks. The Qwen2.5-VL model …
Google has announced several new versions of its Gemini artificial intelligence models, expanding its AI portfolio with different options for developers and users. The company is making Gemini 2.0 Flash generally available through its AI Studio and Vertex AI platforms, while introducing a new cost-efficient model called Gemini 2.0 Flash-Lite and an experimental version of …
European artificial intelligence startups raised €8 billion in venture capital funding during 2024, accounting for 20% of all VC investments in the region. According to a report published by Galion.exe, Revaia, and Chausson Partners, as reported by Romain Dillet for TechCrunch, the UK leads European AI investments, followed by France and Germany. The majority of …
Microsoft’s AI chief Mustafa Suleyman has recruited three senior researchers from Google DeepMind, according to a Financial Times report by Melissa Heikkilä. Marco Tagliasacchi and Zalán Borsos, who developed DeepMind’s podcast-generating “Audio Overviews” feature, will join Microsoft’s new AI lab in Zurich. The third hire, Matthias Minderer, will focus on AI vision capabilities. The researchers …
Google has revised its AI ethics guidelines, eliminating previous restrictions on using AI for weapons and surveillance applications. According to reporting by Nitasha Tiku and Gerrit De Vynck in The Washington Post, the company removed the “Applications we will not pursue” section that had been in place since 2018. Google’s AI chief Demis Hassabis and …
The European Union has implemented its first compliance deadline for the AI Act, allowing regulators to ban AI systems that pose unacceptable risks to individuals. As reported by Kyle Wiggers for TechCrunch, the regulations that took effect on February 2 establish four risk categories for AI applications, with the highest risk category facing complete prohibition. …
Meta AI and the University of Illinois Chicago researchers have created a new system that helps AI models respond more efficiently to queries. As reported by Ben Dickson in VentureBeat, the technology enables AI to determine whether a question needs deep reasoning or a quick response. The system, called Inference Budget-Constrained Policy Optimization (IBPO), uses …
Lightning AI has introduced AI Hub, a new marketplace that enables enterprises to find and deploy AI models and applications within their own cloud environments. As reported by Emilia David in VentureBeat, the platform aims to solve the “last mile problem” in generative AI deployment. The marketplace launches with over 50 APIs, including foundation models …
Amazon Web Services (AWS) is implementing automated reasoning technology to help prevent AI models from generating false information, according to a Wall Street Journal article by Belle Lin. The technology aims to provide mathematical proof that AI responses are accurate within specific domains. AWS’s new tool, called Automated Reasoning Checks, requires customers to establish definitive …