Noetra, backed by 44 Japanese companies, is working on the development of multimodal foundation models. The company aims to provide models that serve as a "mothership" for Japanese companies to build customized AI using their own data.
M-SOLUTIONS has launched a plugin that allows users to centrally manage minutes created with the AI minute-taking tool "LINE WORKS AiNote" as records within the business improvement platform "kintone."
Dentsu Group has introduced generative AI into its accounting approval processes, which handle over 200,000 purchase orders annually. By utilizing Gemini 2.5 Pro to build a system that verifies data consistency and compliance with internal regulations, the company has achieved an accuracy rate of 99%.
China is reportedly considering allowing major tech firms to import Nvidia's RTX Pro 5500 gaming chips, while reports suggest Nvidia CEO Jensen Huang has gained significant influence over President Trump's AI policy and stance toward China.
AI inference infrastructure provider Modal Labs is reportedly closing in on a $750 million funding round led by Accel, marking a valuation increase to $15.75 billion.
A new open-source project enables a cluster of seven ESP32S3 microcontrollers to run a 0.5B parameter LLM using 1.58-bit quantization and high-speed SPI daisy-chain communication.
Ollama v0.35.0 introduces decision models support, allowing users to obtain choices, probabilities, and scores instead of text through a new API endpoint.
One year after its launch, Microsoft Research Asia – Singapore is strengthening its role in Southeast Asia through deep collaborations with academia, government, and industry to translate frontier AI research into real-world applications.
AMD has signed a definitive agreement to acquire spatial intelligence startup World Labs for approximately $8.2 billion in an all-stock deal. Co-founder Fei-Fei Li will join AMD as Executive Vice President and Chief Scientist.
New lightweight models provide fast, calibrated probability outputs for classification tasks, optimized for local execution on hardware like RTX PRO 6000 and Apple M4 Max.
Anthropic engineer Boris Cherny explains the challenges of AI code quality and the distinction between experimental tools and professional production environments.
Anthropic CEO Dario Amodei has introduced a three-step plan to slow AI capability advancement to ensure safety research keeps pace, though critics argue the industry's actions have not matched such rhetoric.
Shopify has expanded its support for WebMCP, allowing browser-based AI agents to navigate checkouts, update delivery details, and submit transactions with buyer authorization.
Climate tech startups are increasingly pivoting to align with the AI wave to secure funding, though some leaders worry the focus on AI-driven data centers may overshadow other critical environmental sectors.
MLX-VLM v0.7.4 introduces support for DeepSeek V4.1, Qwen3.5 MoE, and InternVL3.5, alongside new features for batching and multi-image generation.
OpenAI has introduced a set of preliminary guidelines for implementing structured "safety cases" during frontier reinforcement learning training to mitigate risks.
MicroLLM Lab lets users run seven lightweight large language models in a browser to measure speed and accuracy using WebGPU.