Rafomac News, Tech & Trends That MatterChina is reportedly preparing to ban access to its own free AI models, mirroring recent US restrictions on foreign AI. This development marks a new chapter in the escalating AI cold war between the world’s two biggest tech powers, with breakthroughs and fresh models fueling the fire.
OpenAI Unleashes the Biggest ChatGPT Upgrade Yet
OpenAI has turned ChatGPT into a full-fledged super app, launching three powerful new models collectively dubbed GPT 5.6. This suite includes Solve for complex problems, Terra for daily tasks, and Luna for quick, low-cost jobs. An ultra mode even orchestrates parallel AIs to catch mistakes before delivering results. The sheer power and potential raised such security concerns that the US government delayed the rollout by weeks.
Inside ChatGPT, you can now toggle between work and coding modes. Describe a business task and it pulls data from Gmail, Slack, and Excel to craft a ready-to-use PowerPoint presentation in your corporate template. Coding mode goes further: it reproduces bugs, writes fixes across multiple files, and verifies the solution — all faster than a human engineer could manage in a day.
Voice interaction just got a leap too, with GPT Live offering full-duplex conversations. It listens and talks simultaneously, translating live, cracking jokes, and even passing complex queries to the larger model discreetly for deeper analysis.
This isn’t just a chatbot anymore. ChatGPT can control apps on your computer—updating slides, filling spreadsheets, and running background tasks while you continue working unhindered. It’s the brain, voice, engineer, and assistant rolled into one.
Anthropic’s Claude Hits Your Phone—and China’s AI Ban Rumblings
Not to be outdone, Anthropic brought Claude’s Co-work mode to mobile devices. Now, you can text Claude tasks like prepping meeting documents, drafting emails, scheduling sends, and receiving post-meeting recaps—and it executes, schedules, and follows up seamlessly.
With six months of free Claude Max plans offered to people improving software code worldwide, Anthropic is pressing hard in the AI productivity realm. Yet this expansion collides with geopolitical realities. According to a Reuters exclusive, China’s Commerce Ministry is quietly considering banning access to its leading AI models, including those from Alibaba and ByteDance, for overseas users.
This move looks like payback for the US limiting foreign access to Anthropic’s Fable 5 model. If enacted, it signals a bifurcation where the US and China essentially wall off their most advanced AI tech from each other, potentially shaping who benefits from cutting-edge AI developments.
Elon Musk’s Grok 4.5 and Meta’s Spark 1.1 Raise the Stakes
Elon Musk entered the fray with Grok 4.5, an AI designed especially for coding tasks. Live at x.ai, it can whip up apps like weather dashboards, book review sites, and even 3D cities from simple text prompts. SpaceX is now backing the project, merging rocket science with AI innovation — a combination that founders say delivers powerful, fast, and cheaper alternatives to top-tier AI.
Meanwhile, Meta unveiled Spark 1.1, a multitasking agent AI that can maintain massive codebases, debug by screenshot analysis, and generate Facebook Marketplace listings from product videos. Though Spark excels in tool use and reasoning, it trails Claude’s Opus 4.8 and OpenAI’s GPT 5.5 in software engineering and terminal work.
European Robotics and Image AI Face Off
On the robotics front, French company Mistral rolled out Robostral Navigate, a vision-based robot that navigates office environments without traditional laser sensors. It adapts to humans, waits politely, delivers items, and follows detailed spoken directions. Early days, but it’s a sign of AI-powered robots becoming more autonomous and sensitive in real-world settings.
Meta released Muse Image, an advanced image generation model integrated into its apps, offering features like draw-to-edit and real-time marketplace searches. Meanwhile, ByteDance’s SeaArt 5.0 Pro pushes image creation further, converting sketches to 3D models, editing layers separately, and working fluently across languages.
When tested side-by-side, SeaArt impressed with strict scene consistency in room editing tasks, while OpenAI’s GPT Image 2 favoured creative reinterpretation—even adding unrequested elements like dancers. Both struck a balance between fidelity and imagination, but SeaArt holds the cost edge.
AI Science, Open Source, and Future Predictions
Anthropic launched Claude Science for research, but criticism over its closed nature led to the open-source Open Science, letting users pick any AI model under the hood for free-flowing scientific inquiry. It integrates nearly 300 built-in research tools and preserves user data on personal devices.
A new AI 2040 report from former OpenAI researcher Daniel Kokotajlo offers a hopeful outlook, postponing superintelligence breakthroughs to 2030–2040 with international agreements slowing the race. Without cooperation, though, rapid uncontrollable AI advances remain a looming risk.
Cutting Costs and Raising Efficiency in AI Use
Anthropic also demonstrated how to dramatically cut Fable 5’s operating costs by offloading routine tasks onto cheaper models, using clever task orchestration to preserve most of the performance at less than half the price. The skill now lies in directing AI workloads smartly — blending budget and power like a junior engineer taking smart cues from a senior.
The cloud-based Hermès AI agent platform offers persistent memory and multiplatform access, letting agents work quietly in the background and manage tasks automatically—all without confining users to specific hardware.
Why This AI Cold War Matters
The rapid innovation launches from OpenAI, Anthropic, SpaceX AI, Meta, and ByteDance show a furious pace in AI development. Yet, with China and the US possibly locking down their most powerful tools, the global AI landscape may soon split technologically and ideologically.
The stakes are clear. Whoever controls the fastest, cheapest, and safest AI sets the stage for future digital economies, scientific progress, and military applications. And emerging image, voice, robotics, and research tools hint that AI is already reshaping how we work, create, and solve problems.
Watching China’s tentative AI ban unfold alongside these breakthroughs is essential — this cold war of artificial intelligence may define the rest of this decade.
