Today's Headlines

  • Microsoft makes Windows a home for AI agents as NVIDIA RTX Spark PCs open for preorder
  • Anthropic releases Claude Haiku 5.5, about 75% cheaper to run on average than Haiku 4.5
  • OpenAI extends GPT-6 to free ChatGPT users and introduces Intelligent UI

Today's three stories are about where AI runs and what it costs. Microsoft built into Windows a way to run agents safely on the PC, and opened preorders for new machines built with NVIDIA.

Anthropic made its small model about 75% cheaper on average, and OpenAI is bringing GPT-6, which reached paid plans first, to free ChatGPT users.

Today's Top Three

Microsoft makes Windows a home for AI agents as NVIDIA RTX Spark PCs open for preorder

Microsoft said on October 7 that it is building Windows as the platform for "hybrid intelligence," in which agents run on the PC when that makes sense and reach the cloud when they need to.

The centerpiece is Microsoft Execution Containers (MXC), which is now generally available on Windows 11. Organizations can define which files and networks an agent may touch, and Windows enforces those policies at runtime. Microsoft also showed how Windows identifies which agent acted, and how that activity feeds into Agent 365 and Intune.

Agents that already support MXC include OpenAI's Codex, GitHub Copilot, OpenClaw, Replit, LM Studio and NVIDIA's OpenShell. Anthropic's Claude Code, Box and Perplexity are among those that plan to add support, and Meta's personal agent Muse is coming to Windows as a native app.

On Copilot+ PCs, Copilot will be able, with the user's permission, to read files and recent activity on the PC, take actions across Windows, and call models running locally. Microsoft expects the rollout to begin in the coming months.

On the hardware side, preorders opened for laptops built on NVIDIA RTX Spark, a superchip that pairs a Blackwell RTX GPU with a Grace CPU. It offers up to 128GB of unified memory and up to a petaflop of AI compute. Microsoft's Surface Laptop Ultra starts at $2,599 (MSRP) and goes on sale October 16; in Japan it starts at 513,480 yen, according to ITmedia. Acer, ASUS, Dell, HP, Lenovo, MSI and Gigabyte are also building RTX Spark machines.

At the same event, NVIDIA previewed a Windows version of DGX Station, its deskside AI supercomputer, which until now ran on Linux. NVIDIA says it can run models up to the trillion-parameter scale locally. NVIDIA CEO Jensen Huang and Microsoft CEO Satya Nadella shared the stage to talk about agents working inside the PC.

Anthropic releases Claude Haiku 5.5, about 75% cheaper to run on average than Haiku 4.5

Anthropic released Claude Haiku 5.5, its new small model, on October 7.

The model is aimed at high-volume, cost-sensitive work such as summaries, classification and database queries. Anthropic says it also suits subagent work under Opus 5.5 and Sonnet 5.5, and speed-sensitive jobs like live customer support and browser use. It is the company's fastest model at standard speed, and the first Haiku model with an adjustable effort setting.

For prompts up to 100,000 tokens, Haiku 5.5 costs $0.10 per million input tokens and $0.50 per million output tokens, down from $1 and $5 for Haiku 4.5. Taking into account that it uses slightly more tokens per task, Anthropic says it costs around 75% less to run on average. It is available now on all platforms, including AWS, Google Cloud and Microsoft Azure.

On OSWorld 2.1 (offline subset), which measures how well an agent can operate a real computer through long tasks, Haiku 5.5 scored 72.4%, against 15.7% for Haiku 4.5. Anthropic positions Sonnet 5.5 and Opus 5.5 as the better choices for complex agentic coding.

Anthropic also halved the price of cache reads on Sonnet 5.5, from $0.20 to $0.10 per million tokens, which it says makes Sonnet 5.5 about 20% cheaper on most agentic work. This week it begins giving Max and Team subscribers a monthly API credit: $100 for Max 5x, $200 for Max 20x and up to $500 for Team, pooled across users.

OpenAI extends GPT-6 to free ChatGPT users and introduces Intelligent UI

OpenAI announced on October 7 a GPT-6 model built for the more than 1.2 billion people who use ChatGPT each week, and began rolling it out together with a new capability called Intelligent UI.

With Intelligent UI, ChatGPT composes answers from graphics, tappable buttons, forms, charts and interactive elements, depending on the question. A comparison may come side by side, an explanation as a diagram the user can manipulate, and a simple question as plain text. Users can also ask it to build a small tool on the spot, such as a bill splitter or a savings calculator.

The rollout began on October 7 for Plus, Pro, Business and Enterprise, and extends to the Free and Go tiers the following day. Paid tiers run on GPT-6 Sol, while Free and Go run on GPT-6 Luna. Enterprise availability depends on workplace admin settings.

For questions that need web search, GPT-6 Instant starts answering 44% sooner on average than GPT-5.6 Instant. The update covers the Chat experience; the models powering Work and Codex stay the same. OpenAI introduced its first GPT-6 models for paid customers last month.

Other Developments

Policy

Research

  • OpenAI published a broad set of new mathematical results produced by an internal frontier model in a GitHub repository, with many of the proofs formalized in Lean. It says the average result used compute equivalent to roughly three hours of ChatGPT Pro thinking. Sharing AI progress in mathematics (OpenAI)

Products

Business

Other

Source: Selected by the editors from the AI news inbox (collected October 8, 2026; 29 items, 13 primary and 16 secondary).