Z.ai’s GLM-5.2 beats GPT-5.5 on coding for 1/6th the cost
The open-weights model allows engineering teams to host frontier-level AI on sovereign infrastructure without vendor lock-in.
🤖 AI — 2026-07-11
ddNews · a daily AI briefing for Dirk (generative video/ddStudio, Lit
📰
Krea Releases Open Weights for Image Gen
Enterprise-grade AI image generation is now available as open weights under a custom license for infrastructure control.
Ant Group releases open-source robot brain
LingBot-VLA 2.0 controls 20 distinct hardware types and outperforms pi0.5, advancing cross-embodiment robotics.
Sovereign Inference
DeepInfra opens Canada's first Blackwell-generation inference cluster to address data residency and sovereignty gaps.
GPT-5.6 Security Focus
OpenAI's new model family emphasizes cybersecurity improvements alongside raw capability, addressing recent government concerns.
OpenAI launches GPT-5.6 and ChatGPT Work
OpenAI officially released its most capable model family and a new enterprise agent after US government delays, directly targeting Anthropic on price and speed.
OpenAI GPT-5.6 Launch
OpenAI releases GPT-5.6 with "ChatGPT Work" agent, claiming 54% token efficiency gains on coding while facing regulatory delays.
Security Vulnerabilities
China reports security flaws in Anthropic's Claude Code allowing sensitive data leakage, while Anthropic removes a secret tracker for Chinese users.
Anthropic faces model access restrictions and pricing shifts
The US government is close to restoring access to Claude Fable 5, while Anthropic introduces usage-based fees for top-tier consumer access.
OpenAI launches ChatGPT Work to automate workplace tasks
The new agent combines ChatGPT, Codex, and GPT-5.6 to help white-collar workers access coding tools without high costs.
Meta opens Muse Spark 1.1 to developers
Meta released developer access to its upgraded Muse Spark model, intensifying competition in the generative media space against OpenAI and Anthropic.
Offline-First AI for Healthcare
Experts argue offline AI is the only viable kind for regions with unreliable connectivity like Nigerian healthcare.
Google Photos adds AI video remixing
Users can now stylize and remix personal photo libraries into videos using Gemini Omni, expanding generative video capabilities for consumers.
OpenAI GPT-5.6 Launch
OpenAI officially launches its most capable GPT-5.6 model following US government clearance, marking a major shift in frontier AI availability.
Meta Muse Suite
Meta launches "Muse Image" and "Muse Video" to compete with Midjourney, integrating directly into Instagram workflows.
Cloud Agent Shift
Anthropic moves "Claude Cowork" to the cloud, revealing 90% of sessions are non-coding tasks, while OpenAI pushes "ChatGPT Work" for enterprise automation.
Anthropic expands Claude Cowork to mobile
The AI agent now operates on phones, allowing tasks to continue even after laptops are closed.
Ukraine mandates sovereign AI infrastructure
The Ukrainian government will prioritize AI models that can be run on local servers to ensure digital sovereignty during wartime.
Anthropic Outed for Secret Tracker
The company removed a tracker secretly monitoring Claude Code users in China after a security researcher exposed it.
Hailuo Physics Breakthrough
MiniMax's Hailuo AI video generator gains traction for superior physics simulation in liquid and motion scenarios.
Analysis: Chinese AI models are rapidly closing the gap with U.S. frontier systems
A deep dive into how Chinese offerings are challenging Western dominance in capability and cost.
DeepSeek releases DSpark for faster LLM inference
The new open-source framework claims to speed up decoding by up to 85%, potentially lowering costs for private cloud deployments.
Anthropic launches Claude Science and Slack integration
New tools include a research workbench for scientific workflows and a Slack agent to deepen enterprise adoption.
Nadella Urges Firms to Build Own Models
Microsoft CEO Satya Nadella believes businesses should stop relying solely on frontier models and develop proprietary AI.
China's open-weight models disrupt enterprise economics
Coinbase cut AI spend by 50% using Chinese models like GLM-5.2, which now outperform GPT-5.5 on coding benchmarks for a fraction of the cost.
US AI Security Crackdown
The US government continues to enforce strict access controls on frontier models, recently allowing limited releases of Anthropic's "Mythos" and OpenAI's GPT-5.6 to vetted entities.
Anthropic accuses Alibaba of illicit model extraction
The U.S. AI firm claims Chinese giant Alibaba executed the largest known attack to extract Claude model capabilities.
OpenAI unveils custom chip with Broadcom
The company revealed its first in-house AI chip designed to accelerate infrastructure development and reduce reliance on third-party hardware.
Anthropic Enterprise Push
Anthropic expands its enterprise footprint with "Claude Tag" for Slack and "Claude Science," targeting corporate workflows and research specifically.
Micron secures strategic AI infrastructure deal with Anthropic
The memory giant signed a supply agreement and made a strategic investment to support Anthropic's growing compute needs.