Foundation Models AI News & Updates

Wirestock Raises $23M to Supply Multi-Modal Creative Data to AI Foundation Model Makers

Wirestock, a platform that originally helped photographers sell stock photos, has pivoted to become a data provider for AI labs, raising $23 million in Series A funding. The company now supplies images, videos, design assets, and 3D content from over 700,000 artists and designers to six major foundation model makers, achieving a $40 million annual revenue run-rate. Wirestock focuses on providing high-quality, annotated multi-modal data for creative AI applications like image and video generation.

Genesis AI Unveils GENE-26.5 Foundation Model with Custom Robotic Hands and Data Collection Gloves

Genesis AI has revealed its first foundational robotics model, GENE-26.5, alongside custom-designed robotic hands that match human hand size and shape. The startup has developed a full-stack approach including sensor-loaded gloves for data collection from human workers, simulation systems for rapid iteration, and plans to release a full-body general-purpose robot soon. The company raised $105 million in seed funding and is expanding across Paris, California, and London with a team of 60 people.

Meta Acquires Humanoid Robotics Startup to Advance Embodied AI Research

Meta has acquired Assured Robot Intelligence (ARI), a startup developing foundation models for humanoid robots capable of performing physical labor and adapting to human behaviors. The ARI team, including co-founders Xiaolong Wang and Lerrel Pinto, will join Meta's Superintelligence Labs to advance whole-body humanoid control technology. The acquisition reflects the broader industry belief that achieving AGI may require training AI models through physical world interactions rather than data alone.

Physical Intelligence Unveils Robot AI with Emergent Task Generalization Capability

Physical Intelligence has released research on its π0.7 model, demonstrating that the robot brain can perform tasks it was never explicitly trained on through compositional generalization. The model successfully combined fragmented training data to operate an air fryer and perform other novel tasks, surprising even the researchers who knew the training data intimately. While promising, the system still requires step-by-step verbal coaching for complex tasks and lacks standardized benchmarks for validation.

Agile Robots Partners with Google DeepMind to Integrate Gemini AI Models into Industrial Robotics

Munich-based Agile Robots has entered a strategic partnership with Google DeepMind to integrate Gemini Robotics foundation models into its robots across industrial sectors including manufacturing, automotive, data centers, and logistics. The collaboration will involve testing and deploying AI-powered robots while using data collected from Agile Robots' 20,000+ installed systems to improve DeepMind's underlying AI models. This partnership follows similar deals between Google DeepMind and other robotics companies like Boston Dynamics, reflecting an industry trend toward combining specialized hardware and AI expertise.

Anthropic's Opus 4.6 Achieves Major Leap in Professional Task Performance with 45% Success Rate

Anthropic's newly released Opus 4.6 model achieved nearly 30% accuracy on professional task benchmarks in one-shot trials and 45% with multiple attempts, representing a significant jump from the previous 18.4% state-of-the-art. The model includes new agentic features such as "agent swarms" that appear to enhance multi-step problem-solving capabilities for complex professional tasks like legal work and corporate analysis.

Arcee AI Releases 400B Parameter Open-Source Foundation Model Trinity to Challenge Meta's Llama

Startup Arcee AI has released Trinity, a 400B parameter open-source foundation model trained in six months for $20 million, claiming performance comparable to Meta's Llama 4 Maverick. The model uses a truly open Apache license and is designed to provide U.S. companies with a permanently open alternative to Chinese models and Meta's commercially-restricted Llama. Arcee is positioning itself as a new U.S. AI lab focused on winning developer adoption through best-in-class open-weight models.

Humans& Raises $480M to Build Foundation Model for AI-Powered Team Coordination

Humans&, a startup founded by former employees of Anthropic, Meta, OpenAI, xAI, and Google DeepMind, has raised a $480 million seed round to develop a foundation model focused on social intelligence and team coordination rather than traditional chatbot capabilities. The company plans to build a new model architecture trained using long-horizon and multi-agent reinforcement learning to enable AI systems that can coordinate people, manage group decisions, and serve as connective tissue across organizations. The startup aims to create both the model and product interface together, positioning itself as a coordination layer rather than a plugin for existing collaboration tools.

Skild AI Raises $1.4B at $14B Valuation for General-Purpose Robot Foundation Models

Skild AI, a startup founded in 2023, has raised $1.4 billion in a Series C round led by SoftBank, valuing the company at over $14 billion. The company develops general-purpose foundation models for robots that can be retrofitted to various robots and tasks with minimal additional training, aiming to enable robots to learn by observing humans.

Nvidia Launches Comprehensive Physical AI Platform for Generalist Robotics at CES 2026

Nvidia unveiled a complete ecosystem for physical AI at CES 2026, including robot foundation models (Cosmos Transfer/Predict 2.5, Cosmos Reason 2, Isaac GR00T N1.6), simulation tools (Isaac Lab-Arena), and new Blackwell-powered Jetson T4000 edge hardware. The company aims to become the default platform for generalist robotics development, similar to Android's dominance in smartphones, by making robot training more accessible through partnerships with Hugging Face and offering open-source tools. Major robotics companies including Boston Dynamics, Caterpillar, and NEURA Robotics are already adopting Nvidia's technology.