Google is rolling out Gemini 4 Argon, its new flagship AI model, after delays forced the company to scrap an earlier version called Gemini 3.5 Pro. The model performs well on benchmarks in legal reasoning, finance, and cybersecurity, though it is not yet available to the public and has no release date. Google is pricing it at $2 per million input tokens and $10 per million output tokens, roughly half the cost of some rivals. Google Cloud revenue grew 82% year over year to $24.8 billion in Q2, and the company plans heavy data center spending of $195 billion to $205 billion in 2026.
Key Takeaways
- Google launched Gemini 4 Argon, competing with OpenAI and Anthropic, at about half the cost of some rivals
- Palantir partnered with Armada to deploy sovereign AI data centers in months, not years
- New research shows small AI models can team up with large models under budget constraints
- A federal judge warned AI could harm lawyer training and client outcomes
- Cerebras shares fell 20% as Nvidia's dominance in AI chips persists
- Hugging Face, Anthropic, Google, DeepLearning.AI, and Stanford offer free AI agent courses for 2026
- SoundHound AI reported record Q2 revenue of $61.9 million despite stock pressure
- Charles Wu outlined five AI investment edge paths for investors
- AI agent traffic growth is boosting internet infrastructure stocks like Cloudflare and Akamai
- Majan University College uses AI tools to reshape cybersecurity teaching
Google launches Gemini 4 Argon to compete with OpenAI and Anthropic
Google released its new flagship AI model called Gemini 4 Argon. The model performs well on benchmarks in legal reasoning, finance, and cybersecurity. Google is rolling it out slowly, starting with cybersecurity partners and the U.S. government. The model is already being used inside Google to save memory in its data centers.
Alphabet launches Gemini 4 Argon after delays and focuses on low price
Alphabet unveiled Gemini 4 Argon after delays that forced Google to scrap an earlier model called Gemini 3.5 Pro. The model is not yet available to the public and has no release date. Google is pricing it at $2 per million input tokens and $10 per million output tokens, about half the cost of some rivals. Google Cloud revenue grew 82% year over year to $24.8 billion in Q2, and the company plans heavy data center spending of $195 billion to $205 billion in 2026.
Palantir and Armada partner to build AI data centers in months
Palantir and Armada announced a partnership to deliver sovereign AI infrastructure. The deal combines Armada's Galleon modular data centers with Palantir's Sovereign AI Operating System. Customers can deploy AI models on infrastructure they own and control, built in the United States and allied nations. The setup can be deployed in months instead of years, and the financial terms were not disclosed.
Palantir shares rise after AI infrastructure deal with Armada
Palantir shares closed up 1.6% at $190.04 after announcing its partnership with Armada. Armada was named Palantir's first Certified Modular Data Center Partner. The deal gives Palantir a physical infrastructure layer for AI systems that can run on-premises, at the edge, or in sovereign environments. No financial details of the agreement were revealed.
New study shows how small and large AI models can work together
Researchers developed a way for small language models to team up with large language models under a budget. The small model stays in charge of reasoning and only asks the large model for help when needed. They built a three-stage RLVR framework that teaches the small model how to query and use the advisor. The approach improves performance versus cost on math reasoning and coding tasks.
SoundHound AI stock drops 3.2% amid integration concerns
SoundHound AI fell 3.2% due to pressure on AI stocks and concerns about its LivePerson acquisition. The company completed the acquisition in September and appointed a new CFO. Recent SEC filings show share resale registration, which may worry investors. However, SoundHound reported record second-quarter revenue of $61.9 million and raised its full-year outlook.
Charles Wu outlines five AI investment edge paths
Charles Wu, director at Frontier Advisors, identified five ways investors can gain an edge using AI. These include information edge, speed edge, investment horizon edge, implementation edge, and structural balance sheet edge. He noted that AI can reduce transaction costs and increase coverage for managers. Wu emphasized there is no single approach that works for all investors.
Judge warns AI may harm lawyer training and clients
A federal judge warned that law firms using AI could risk harming clients and slowing the training of junior lawyers. Manhattan US District Judge Arun Subramanian made the warning in an order this week. He declined to sanction attorneys over AI-generated errors in a court filing. He used the episode to highlight broader concerns about AI in legal practice.
AI agent traffic boom boosts internet infrastructure stocks
AI agents are driving major web traffic growth, benefiting internet infrastructure companies. Jim Cramer highlighted Cloudflare, Akamai, and Fastly as key players but recommends only one as a buy. AI agents operate continuously and generate large data requests, straining current infrastructure. This shift may create more stable revenue models for companies that capture AI traffic.
AI-powered gamified method reshapes cybersecurity teaching
A cybersecurity instructor at Majan University College uses AI as a core teaching tool to build critical thinking. Students use Tenable's Nessus Essentials to find real vulnerabilities on their own machines. They then compare remediation advice from ChatGPT, Google Gemini, and Claude AI. This approach helps students apply knowledge in practical scenarios and debate different AI-generated solutions.
5 Free AI Agent Courses to Get Started in 2026
Five free online courses can help you learn how to build AI agents. Hugging Face, Anthropic, DeepLearning.AI, Google, and Stanford University each offer a course covering topics like transformer models, AI safety, machine learning basics, and neural networks for image recognition. All courses include video lectures, quizzes, and hands-on exercises. They are available online at no cost to anyone interested in AI development.
Cerebras Stock Falls to Post-IPO Low as Nvidia Pressure Mounts
Cerebras Systems shares dropped 20% to their lowest point since going public due to pressure from industry leader Nvidia and insider selling after lockup expiration. The company built its reputation on the CS-2 system, which features the world's largest computer chip with 2.6 trillion transistors designed for AI workloads. Despite its technical innovation, Cerebras struggles to gain market share because Nvidia's CUDA software platform has become the standard for AI development. Public market investors are growing skeptical about AI chip startups trying to challenge Nvidia's dominance.
Sources
- Google has been playing catch up with OpenAI and Anthropic. Does its new flagship model really compete at the frontier?
- Alphabet Unveils Gemini 4 Argon After Delays, and Bets on Price
- Palantir and Armada say their AI data centers can be deployed in months, not years
- Palantir Shares Climb After New AI Infrastructure Deal With Armada
- Learning to Ask: Information Acquisition for SLM-LLM Collaboration, under a budget
- Why SoundHound AI Stock Fell Today
- Frontier’s Charles Wu maps five paths to turning AI into an investment edge
- Judge warns AI could stunt lawyers' training and harm their clients
- AI Agents Surge Drives Internet Infrastructure Stock Rally
- Teaching Cyber Threats Differently: An AI-Powered Gamified Approach
- 5 Free AI Agent Courses to Learn How to Build Agents in 2026
- Cerebras Stock Crashes to Post-IPO Low Amid Nvidia Pressure
Comments
Please log in to post a comment.