AI Token Pricing Hit New Record Lows In a Competitive Landscape

AI token pricing hit new record lows last month in an increasingly competitive landscape. Compared to the high recorded earlier this summer, it represents a fall of more than half. A key gauge of daily pricing, the LLM Token Expenditure Index, developed by AI firm Silicon Data, fell to 97 cents on Monday. The recent drop can be bad news for the companies behind the models, such as OpenAI, Anthropic, and Google’s Gemini. The decrease is driven in part by the rise of open-source Chinese models like Moonshot’s Kimi K3. In response, American leading companies are announcing price cuts.

Google Introduces Gemini 3.8, Its Best Reasoning and Coding Model

Nvidia's CEO Says That The Era of AGI Has Arrived With GPT-6 Astra

NYC's Department of Education Places Strict Limits On AI Tool Usage in Classrooms

Hugging Face Approached Nvidia Weeks Ahead of the Now Confirmed $12.9B Acquisition

Moving beyond hardware and further up the AI stack, Nvidia last Thursday officially agreed to buy open-source platform Hugging Face for $12.9 billion. This is Nvidia’s second-biggest, following the $20 billion purchase of assets from chipmaker Groq in December. Before that, its largest deal was the purchase of Israeli chipmaker Mellanox for almost $7 billion in 2019. Nvidia's stated goal is to keep Hugging Face as an open platform for the entire AI ecosystem, Nvidia CEO Jensen Huang wrote in a blog post Thursday. "Together, we will scale Hugging Face’s platform, strengthen its infrastructure and expand access to AI for developers and institutions worldwide,” he wrote. Hugging Face CEO Clément Delangue revealed that he recently approached Huang about a deal, thinking that Nvidia was "a perfect home for his company." “We realized that Hugging Face and open-source AI in general were at the turning point, and that it needed more, more resources, more scale, more visibility,” he said. Nvidia has become the world’s most valuable company due to the growing demand for its graphics processing units, which power the AI boom.  

OpenAI Introduces GPT‑6 Astra, Saying That With This Model "We Are Now In The AGI Era”

OpenAI announced GPT‑6 Astra yesterday, which it labeled "the world’s most intelligent and aligned model," as "it brings together years of research and big bets across pre-training, reinforcement learning, and alignment." OpenAI president Greg Brockman said that with this model, "we are now in the AGI era.” GPT‑6 Astra was rolled out to a limited set of organizations in the cybersecurity industry as a preliminary step toward making it available to all ChatGPT Plus, Pro, Business, and Enterprise users, as well as through the OpenAI API, Microsoft Azure, and AWS Bedrock. "Astra is our most aligned model, with substantial improvements in understanding user intent and model behavior—you can delegate tasks with greater confidence in Astra’s judgment," said the San Francisco – based lab. OpenAI stated that GPT-6 Astra would have avoided the Hugging Face hacking episode. GPT‑6 Astra can handle tedious tasks such as filling out online forms, updating customer records in a CRM, creating a website, and installing and testing software. It’s also the first model designated as meeting OpenAI’s “critical cybersecurity capability threshold” — but the company promises that won’t lead to a repeat of its model hacking a rival company’s internal systems. "It's a generational leap in capability for areas like cybersecurity, professional work, software engineering, science, and computer use." Notably, the API cost for the new model will be $10 per million input tokens and $50 per million output tokens. That's the exact same price as Anthropic's Mythos/Fable 5 and 5.1, and makes Astra one of the most expensive models on the market. However, OpenAI emphasizes that Astra's leap in intelligence causes it to use "substantially fewer total tokens per task" in multiple scenarios. So this is also an efficiency play from that perspective. In real-world usage, we'll have to see how well that translates to lower overall costs. This new model comes a year after the release of GPT-5 and nearly two months after the release of GPT-5.6, the last iteration of the previous model suite. OpenAI touted GPT‑6 Astra agentic capabilities in a bid to attract enterprise customers — and compete with Anthropic ahead of its IPO. • Demo Videos of GPT-6 Astra

OpenAI Announces That It's Solving Unanswered Math Questions

Now All OpenAI’s Paid Users Have Access to GPT-6 Astra

Anthropic Introduced Claude Fable 5.1 and Claude Mythos 5.1., Its Most Advanced AI Models

Alibaba Releases Qwen3.8-Flash, a 125 Billion Parameters Open-Source Model

Alibaba Group released the latest model in its popular Qwen series, Qwen3.8-Flash, which is available for download starting August 26. With 125 billion parameters, its performance is competitive with some of the latest releases by rivals such as Anthropic’s Opus 4.6 and DeepSeek’s V4-Flash, Alibaba said in a statement. Qwen3.8-Flash uses an architecture designed for the next generation of models: the Qwen 4 series. That new system significantly reduces training and inference costs. Alibaba is China’s biggest spender on AI, pledging more than $57 billion over three years toward AI development and infrastructure building. To fund that effort, it raised US$10.2 billion last month in Hong Kong’s largest follow-on share offering.

Google Launches Gemini Enterprise for Legal, With New Tools for Lawyers and Law Firms

Google expanded its Gemini Enterprise AI platform last month with new tools for lawyers and law firms, joining the race among tech companies in the legal sector. According to the company, the new Gemini Enterprise for Legal plug-in will help law firms use a suite of AI agents to manage both routine and complex work, including legal research, drafting filings, client and case management, and other legal tasks. And will integrate with legal software and data platforms, including Thomson Reuters, Harvey, Legora, and others. Alphabet's Google launched Gemini Enterprise for business customers in ⁠October. A 2026 report on the legal industry's adoption of AI from Deloitte found that law firms are investing heavily in the technology as part of a transformation that's still in its early stages. • Anthropic and other rival companies have also expanded their enterprise offerings in recent months • Thomson Reuters, which owns Reuters and the Westlaw legal research database, recently launched a proprietary LLM called Thomson 1.0, trained on ‌its legal ⁠research content and built for professional work. It said the model is built on open-source technology and draws on decades of its own "authoritative" legal content. • Some other large law firms have formed strategic partnerships with AI providers and, in some instances, are building their own AI platforms. Google said Weil Gotshal, Cleary Gottlieb, Freshfields, and Williams & Connolly are among the firms collaborating with it on Gemini Enterprise for Legal.

......

The AI-Native University

Blueprint for the Institution of the Future

A practical guide for leaders reimagining higher education with AI at the core — not as an add-on, but as the foundation for how students learn, faculty teach, and institutions govern.

By Mikel Amigot

Read the book