Daily AI brief · October 8, 2026
GPT-6 Rolls Out to ChatGPT's 1.2 Billion Weekly Users
OpenAI made GPT-6 the default in ChatGPT with answers that include buttons, forms and charts, Anthropic cut small-model prices with Claude Haiku 5.5, and Microsoft priced its local-AI Surface at $2,599.
GPT-6 becomes the default in ChatGPT, with answers you can click (OpenAI, BusinessToday, gHacks) OpenAI is rolling GPT-6 out to everyone who uses ChatGPT, which it says is more than 1.2 billion people a week. Plus, Pro, Business and Enterprise accounts get the larger GPT-6 Sol model now, and Free and Go accounts get GPT-6 Luna starting a day later. It arrives with "Intelligent UI": instead of a wall of text, an answer can include buttons, forms, charts, diagrams and small working calculators that appear as the reply is written. For a small business, the practical part is that the tool your staff use every day just changed under them, so the prompts, templates and checklists that worked last week deserve a quick rerun before anyone sends the results to a customer.
Anthropic cuts the price of its small model with Claude Haiku 5.5 (VentureBeat, MarkTechPost, Neowin) Anthropic released Claude Haiku 5.5 at $0.10 per million input tokens and $0.50 per million output tokens for shorter requests, with a context window of one million tokens and higher rates for very long prompts. Reviewers point out the new model uses more tokens for the same job, and Anthropic's own estimate is that real workloads cost about 75% less than on Haiku 4.5 once that is counted. If someone in your business has been holding off on an automation like sorting the inbox, tagging support tickets or summarizing call notes because of cost, the math just moved in your favor, but judge it on a week of your own bills, not the headline rate.
Microsoft prices its local-AI laptop at $2,599 and gives Copilot a Windows overhaul (Engadget, The Gadgeteer, Windows Report) At yesterday's Windows and Surface event, Microsoft showed the Surface Laptop Ultra, built on Nvidia's RTX Spark hardware and able to run AI models of around 120 billion parameters without the cloud. It starts at $2,599, with partner RTX Spark PCs also shipping October 16. Microsoft also redesigned the Copilot app with Home, Code and Autopilot sections that can reach local files and take actions in Windows, and announced Microsoft Execution Containers, a way for developers to fence in what an AI agent can touch. Whoever buys your computers can relax: this is a premium tier for people with heavy AI work, and the more important news for most offices is that Copilot is getting permission to act on your files, which means someone should decide who turns that on.
Mistral previews a 1 trillion parameter model you can run yourself (MarkTechPost, The Next Web) Mistral, the French AI company, opened a preview of Mistral Large 4 through its own service and says the downloadable version arrives on October 27, so organizations can run it on their own servers under their own rules. It handles text and images, and Mistral trained it in its own European data centers. A model this size is beyond a typical office server, but if you work in a field where client data cannot go to an outside AI company, this is one more sign that a private option through your IT partner is becoming realistic.
Retest your top three AI tasks. Have your AI champion pick the three things your team asks ChatGPT most often, run them again on GPT-6, and compare the answers with last week's. Look for changed numbers, missing steps and new interactive pieces that don't copy cleanly into an email or document.
Add one line to your staff AI rules. "Copilot or any AI assistant may not be given access to company files or allowed to take actions on a computer until the owner approves it." Microsoft is building those switches into Windows now, and it is easier to set the rule before the switch shows up.
Skip the $2,599 laptop for now. Unless someone on your team is already running AI models on their own machine, the new Surface is more computer than an office needs. Wait for the reviews and for prices on partner machines.
Where it's going: The price of everyday AI keeps falling while the big tools change without notice, so the advantage goes to the businesses that write down what they use AI for and check it whenever the ground shifts.