Latest news, developments, and analysis from the world of artificial intelligence. Up-to-date information about AI technologies.
Prism introduces a free, unified LaTeX workspace designed to accelerate everyday scientific work by integrating GPT-5.2 directly into the drafting process. With features like project-aware AI, unlimit
Google has integrated DeepMind's Lyria 3 model into Gemini, allowing users to generate 30-second songs with vocals from text or images. Featuring SynthID watermarking for safety and custom cover art b
Google introduces Gemma 4, delivering unprecedented intelligence-per-parameter under an open Apache 2.0 license. The 31B model ranks #3 globally on Arena AI, while new multimodal edge variants redefin
Mercedes-Benz combines its automotive legacy with NVIDIA's AI expertise. The new S-Class features a Level 4-ready architecture and Uber integration, redefining safety and luxury mobility.
NVIDIA democratizes weather forecasting with the fully open "Earth-2" AI model family, accelerating global climate analysis from initial data to local storm predictions.
Alibaba Cloud unveils Qwen3.6-Plus, establishing a new global standard for autonomous AI agents. Featuring a 1M context window and scoring an unprecedented 78.8 on SWE-bench Verified, it flawlessly ex
Researchers have unveiled MapTrace, a scalable pipeline utilizing Gemini 2.5 and Imagen-4 to generate 2 million synthetic maps. This breakthrough teaches MLLMs fine-grained spatial reasoning, drastica
Qwen3-Max-Thinking sets a new standard with advanced reasoning, adaptive tool use, and performance that surpasses Gemini 3 Pro in key benchmarks.
Introducing Kimi K2.5, the most powerful open-source model trained on 15T tokens. Featuring a native multimodal architecture and the groundbreaking "Agent Swarm" paradigm, K2.5 can orchestrate 100 sub
Using Meta’s AI models, autonomous robots developed for a DARPA challenge are transforming how medics prioritize casualties in disaster zones.
Google DeepMind unveils "Agentic Vision" for Gemini 3 Flash, transforming AI image analysis from a static snapshot into an active investigation. By executing Python code to zoom, annotate, and analyze
The Qwen3-TTS family is now open-source, offering powerful voice cloning, custom voice design via prompts, and ultra-low latency streaming capabilities for developers worldwide.
Google has launched Project Genie, an experimental research prototype powered by the Genie 3 world model, for Google AI Ultra subscribers in the U.S. This groundbreaking platform allows users to sketc
Google Labs has officially acquired ProducerAI, a generative music platform powered by Lyria 3 and Gemini. Offering SynthID protection, the tool allows artists to create full tracks and custom instrum
Alibaba Cloud unveils Qwen3.5-397B-A17B, an open-weight model featuring a revolutionary hybrid architecture. Activating only 17B of its 397B parameters for efficiency, it outperforms GPT-5.2 in LongBe
Google introduces Nano Banana 2, blending advanced world knowledge with Gemini Flash speeds. The model boasts 4K resolution, precise text rendering, and character consistency for up to five subjects,
Microsoft unveils Maia 200, a 3nm inference accelerator powering GPT-5.2 with industry-leading efficiency, delivering 3x the performance of Amazon Trainium3.
Revolutionizing neuroscience, TRIBE v2 is a predictive foundation model offering a 70x resolution increase to simulate human neural responses. Released entirely open-source, it empowers clinical break
Scientists are treating Large Language Models not as code, but as biological entities to unlock their secrets. This new approach reveals the alien "mind" behind the 200-billion-parameter giants.
Driven by clear ROI, financial institutions are scaling up AI budgets. The latest NVIDIA survey highlights a strategic shift towards open-source models and autonomous AI agents to drive growth.