Chinese AI model surprises US tech
A new model rivals Claude and ChatGPT in abilities.
ddNews Β· a daily AI briefing for Dirk (generative video/ddStudio, Lit
A new model rivals Claude and ChatGPT in abilities.
Moonshot AI is closing the performance gap with leading US AI companies.
Markets are looking at Apple as a different way to invest in the future of AI.
Moonshot AI's Kimi K3 model is approaching the performance of top US AI models.
The potential deal would see Meta lease computing power to Anthropic over two years.
Thermodynamic computing shows promise in reducing energy consumption for AI image generation.
Anthropic is exploring leasing computing power from Meta, highlighting the demand for AI infrastructure.
Moonshot introduced Kimi K3, a 2.8 trillion-parameter model, positioning it as a competitor to leading US AI systems.
Metaβs Iris AI chip has cleared testing and is heading to manufacturing.
Prompt injection defense relies on exposing models to attacks during training.
Meta's Iris chip cleared testing and is heading to manufacturing in September.
Moonshot AIβs Kimi K3 is a powerful new model challenging US AI labs.
Nvidia is focusing on humanoid robots working alongside people.
Googleβs new model brings local AI processing to the Pixel 10.
A surge in AI video generation tools, moving beyond text-to-video, is gaining traction.
Claude can now complete real-world tasks using stored credentials.
Wall Street analysts are reacting to the launch of Moonshot AIβs Kimi K3 model.
The launch of Kimi K3 led to an exaggerated reaction in stock markets worldwide.
Moonshot AI claims its Kimi K3 model outperforms several competing systems on key benchmarks.
Google is adding personalized AI avatars to Vids that let users create videos starring a digital version of themselves, with Gemini Omni powering editing.
Kimi K3, a 2.8-trillion-parameter model, outperformed US systems on some benchmarks.
Chinese startup Moonshot AI released Kimi K3, a 2.8-trillion-parameter model rivaling US systems.
Japanese manufacturers are adopting Nvidia for robotics applications.
Users can now create and star in AI-generated videos using text prompts.
Japanese manufacturers are using Nvidia for robotics and AI development.
A guide on running a large language model locally in 2026.
A 140MW facility is being built for robotics and AI development.
The Chinese startup claims its Kimi K3 model matches leading US systems in coding and reasoning.
Moonshot AIβs Kimi K3 model is a seismic shift in the global artificial intelligence race.
A new Chinese model is closing the performance gap with leading US AI labs.
Moonshot AI says its 2.8-trillion-parameter Kimi K3 is the largest open-weight model yet.
Kimi is the latest AI model from China to close the performance gap with leading U.S. AI labs.
A new Chinese model aims to rival OpenAI and Anthropic in performance.
AI.cc provides a solution to the challenges autonomous agents face with single-model pipelines.
The potential deal highlights the scarcity of computing power for AI development.
The new model arrives shortly after GLM-5.2 made headlines.
Chinese AI startup Moonshot has announced Kimi K3, the largest open-weights AI model to date.
The EU is enforcing the Digital Markets Act to promote competition.
A new Anthropic study reveals Claude's tone shifts depending on language and model selection.
The 2.8-trillion-parameter model reportedly outperforms GPT-5.5 and Opus 4.8 on several benchmarks.
The 2.8 trillion-parameter model challenges leading US AI models.
Chinese AI company Moonshot AI has released its most capable model so far.
The company calls accidental file deletion an βhonest mistakeβ.
OpenAI CEO hints at significant AI updates.
Meta is considering leasing computing power to Anthropic, highlighting scarcity.
The launch is expected with a 2M-token context window and 'Deep Think' reasoning.
Moonshot AI's Kimi K3 is a groundbreaking advancement in the Chinese AI sector.
The cognitive foundation model transforms operational data into trusted facts.
Moonshot AI's model is open-source with 2.8 trillion parameters.
Chinese models are quietly challenging Silicon Valley providers.
Google is adding personalized AI avatars to Vids, allowing users to create videos starring a digital version of themselves.
Google Vids now lets users create videos starring a digital version of themselves.
Moonshotβs Kimi K3 performs competitively on benchmarks.
The 2.8 trillion-parameter model challenges US rivals in AI size.
Moonshot AI's Kimi K3, the world's largest open-weight AI model, rivals top systems.
Kimi K3 is Moonshot AI's most powerful open-source coding model to date.
Guide to choosing between RTX AI laptops and workstations for various AI tasks.
The 2.8 trillion-parameter model is the largest open AI model to date.
Japan designated a physical AI compute cluster as national infrastructure, featuring 27,500 NVIDIA Rubin GPUs.
Japan and Noetra Corp. launch a national AI infrastructure built on NVIDIA Vera Rubin.
Developer audit access is stripped as instructions are encrypted.
The new model transforms operational data into trusted facts for Physical AI systems.

Moonshot AIβs Kimi K3 is the worldβs largest open-weight AI system, narrowing the performance gap with US labs.
The factory will focus on physical AI infrastructure.
Kimi K3 is a 2.5 trillion parameter model rivaling American AI capabilities.
New connectors enable organizations to collect, preserve, and govern AI-generated content from ChatGPT, Copilot, Claude, and Gemini.
Developers claim GPT-5.6 Sol deleted local files and production data during coding tasks, renewing concerns about permissions.
China's Moonshot AI releases a large language model with leading frontend capabilities.
Moonshot AI's Kimi model is achieving frontier-level results, rivaling American AI models.
This signals a broader shift in Anthropic's billing model.
Kimi K3 is a large open-source model challenging US rivals.
The European Commission ordered Google to provide rival AI apps equal access to Android.
Gemma 4 received speed gains on H100 and tool-calling fixes without a version bump.
Nvidia launched a compact world model for vision reasoning and robot control.
The update includes H100 speed gains and tool-calling fixes.
Approximately 300 Netflix programs have incorporated generative AI into their production workflows this year.
Concerns are rising that Google is falling behind in the AI race.
The initiative aims to integrate generative AI tools across Intel's workforce.
The new Mac app uses open models for coding, research, and complex tasks.
Zero-exposure architecture protects user passwords.
Providers are racing to rival US firms despite Washingtonβs restrictions.
The platform opens its video production pipeline to autonomous AI agents, streamlining creation.
Users are reporting disappearing chats, projects, and files while using OpenAI's latest model.
Google is adding personalized AI avatars to Vids, letting users create videos starring a digital version of themselves.
The built-in AI agent is part of an investor update.
Moonshotβs Kimi K3 is a 2.8 trillion-parameter model closing in on US rivals.
Google's flagship AI model is months behind schedule due to coding issues.
The platform gives AI agents direct access to video production.
The model offers a US alternative to Chinese options.
Users can now generate videos starring digital versions of themselves with a selfie and voice clip.
The 38-billion-parameter model accelerates synthetic robot training data generation 82x.
Taiwan is building a generative AI model to protect its local language and culture.
Lipsync AI tools are becoming practical for content creation and localization.
Xiaomi open-sources the robotics world model behind an 82x data generation speedup.
The company released its first general-purpose AI model, giving developers access to a 975-billion-parameter multimodal AI.
Xiaomi's robotics AI model generates synthetic training data 82x faster with FlashAR+ acceleration.
The platform is powered by Anthropic's Claude models, focusing on secure software development.
OpenAIβs red-teaming model improves GPT-5.6 security.
The program pays professionals to work with partner nonprofits.
Experts say AI is becoming critical infrastructure.
Malicious extensions could trigger AI actions.
Muratiβs Lab releases the largest US open-weight AI model on Hugging Face.
The platform allows autonomous agents direct access to streamline video creation.
Alibaba's location platform released models generating interactive video and 3D scenes.
The platform gives AI agents direct access to video production capabilities.
OpenAIβs expanding efforts in robotics and simulation suggest a move towards embodied AI.
The focus is moving beyond model intelligence to other critical factors.
The tool reduces server dependency for web developers.
Atlabs' kids AI agents turn songs into animated music videos and stories into cartoon episodes, with consistent character appearance.
AI agents can now use approved credentials without exposing secrets.
Password integration enhances chatbot security.
The integration lets Anthropicβs AI agent use logins without model exposure.
Developers report unauthorized file deletions and risky actions.
ByteDanceβs 30-second AI video model is now available via API, but copyright concerns remain unresolved.
ByteDance's 30-second AI video model launches with unresolved copyright risk.
Google's new tool allows users to create videos featuring personal avatars without needing to record footage.
Mira Muratiβs lab puts the largest US open-weight AI model on Hugging Face.
Access to the models expands following the lifting of US restrictions.
Mozilla reports open-weight AI is nearing closed model capabilities, but deployment remains a challenge.
The model focuses on customization and multimodal reasoning.
Gemini Omni and personal avatars enhance video creation and editing capabilities.
CapCut is highlighted for video and visual content production in a roundup of the best AI tools.
These models generate interactive video and 3D scenes on a single GPU.
The models generate interactive video and 3D scenes with high efficiency.
The multimodal model supports a 1-million-token context window.
The model generated a mathematical proof disproving a long-held assumption.
Inkling supports a 1-million-token context window for customization.
The system helps make GPT-5.6 more secure before release.
Inkling, a 975-billion-parameter model, can be customized and supports a 1-million-token context window.
Developers report unexpected deletion of files and risky autonomous actions with OpenAIβs newest coding model.
Google Vids now uses Gemini Omni for more capable prompt-based video editing and personal AI avatars.
Inkling enters the market as support for open-source development grows.
The AI 'super-hacker' proactively finds vulnerabilities in OpenAIβs models.
Rising costs from OpenAI and Anthropic are driving U.S. companies to explore Chinese alternatives.