Welcome to the latest edition of The Neural Link!
It’s been a busy month in the world of AI model development and the pace isn’t slowing. From open-source breakthroughs to flagship releases, major players across the globe have unveiled their latest models. Here’s a roundup of the most significant launches shaping the future of generative AI.
|
Model |
Company |
Type |
Notable Features |
|
OpenAI |
Flagship general-purpose model |
Released to all ChatGPT users; faster, smarter, "PhD-level" reasoning capabilities |
|
|
|
Open-source AI agent (command-line tool) |
Enables developers to interact with AI models via the command line. Built to simplify building and testing AI workflows and agents locally. |
|
|
Moonshot AI |
Open-source LLM (1T params) |
Strong coding and agentic abilities; aims to reclaim domestic market edge |
|
|
Zhipu |
Open-source "agent-native" LLM |
Reasoning, coding, autonomous tool use; positioned as low-cost alternative |
|
|
Alibaba |
Open-source coding model |
Outperforms domestic rivals; rivals Claude and GPT-4 in coding tasks |
|
|
Mistral |
Open-weight audio model |
Transcribes 30–40 min audio; low-cost and locally deployable |
|
|
OpenAI |
Open-weight reasoning models |
Freely downloadable; run on modest hardware; first open-weight models since GPT-2 |
|
|
xAI |
General-purpose LLM |
Solves complex tasks with agent-like tool use; includes heavy mode for tough problems |
Want to go from theory to action? Learn about AI agents in your workplace.
Join our upcoming webinar on Copilot Agents to explore how Microsoft’s latest tools will reshape workflows, from automation to decision-making. We’ll break down real-world use cases, architecture, and how to build your first agent.
While tools like Copilot Agents are transforming the way we work, the broader AI landscape is evolving just as fast.
Here are the key stories shaping AI across business, policy, and innovation this month.
🕵️OpenAI launches ChatGPT “agent” mode
OpenAI introduced a general-purpose ChatGPT agent that can autonomously navigate apps, browse websites, and execute code on users’ behalf. Rolling out to paid subscribers, this Agent mode marks OpenAI’s boldest step toward AI that “takes actions” for users, not just chats.
📝Trump unveils US AI policy via executive orders
President Trump issued a sweeping AI Action Plan along with multiple executive orders aiming to spur innovation on U.S. terms. The agenda favors letting AI firms “make their own rulebook” and bans “woke” AI in federal use, but critics warn the ideological bent could constrain development and speech.
🔬Stanford “virtual scientists” autonomously tackle biomedical research
Stanford researchers built an AI-powered virtual lab with an “AI principal investigator” and several agent collaborators (immunologist, ML expert, critic, etc.) that brainstorm and experiment together. In a test, the AI team devised a novel nanobody-based COVID vaccine concept within days, going beyond prior human ideas – hinting at a future where AI agents accelerate scientific discovery.
🌏DeepMind maps Earth with AI
Google DeepMind introduced AlphaEarth Foundations, an AI model that fuses satellite imagery, radar and climate data to map Earth’s ecosystems in unprecedented detail. Scientists say this “virtual satellite” model will help identify uncharted habitats and focus conservation efforts.
📈Anthropic overtakes OpenAI in enterprise share
A Menlo Ventures report finds Anthropic’s AI models now hold 32% of enterprise LLM usage, vs. OpenAI’s 25%, marking a rapid reversal since 2023. Claude models are especially dominant in coding tasks, doubling OpenAI’s share in that domain.
- US export gridlock freezes Nvidia’s China chip orders, with billions in AI hardware stalled by license delays at the Commerce Department.
- Yahoo Japan mandates generative AI for all staff by 2028, targeting a 2× productivity boost through tools like SeekAI.
- Zuckerberg unveils “personal superintelligence” vision, pushing Meta’s AGI ambition and acknowledging a fierce talent war for AI researchers.
- GitHub Copilot tops 20 million users, with rapid enterprise growth and widespread adoption across the Fortune 100.
- Netflix slashes VFX costs with AI, generating scenes 10× faster using generative tools in its sci-fi series El Eternauta.
- Google’s Big Sleep AI agent blocks live cyberattack, marking a leap in autonomous, real-time cyber defence.
- Elon Musk’s xAI demo backfires, as Grok 4 goes off-script with offensive content, forcing a public reset.
- Replit AI agent deletes a live database, then falsely reports success—prompting calls for stronger oversight of autonomous coding bots.
That’s a wrap on this month’s AI update. The landscape’s shifting fast and so are the possibilities.
We’ll be back soon with more.
Matt Dunn
Head of AI & Automation
Want more? Contact us to discover how you can partner with AI and automation experts who understand your business needs and industry challenges. You can also subscribe to The Neural Link, our monthly newsletter filled with the latest trends and insights in AI and automation.
