AI news for Thursday, October 8, 2026: OpenAI brings GPT-6 to every ChatGPT plan, Anthropic cuts Haiku prices

GPT-6 reaches every ChatGPT plan and small models get cheap, while a teen safety report and a Windows agent sandbox show who is minding the guardrails.

Models Products3 min
OpenAI brings GPT-6 to every ChatGPT plan, with clickable answers
OpenAI is rolling GPT-6 Sol and Luna into ChatGPT on free and paid plans, with answers that arrive as charts, forms and buttons. The model number is the smaller half of this launch.
Why it matters If you publish calculators, explainers or simple web tools, ChatGPT now builds them inside the answer. If you roll ChatGPT out to a team or a family, read OpenAI's own safety notes first: they list regressions, including on its under-18 tests.
SRCOpenAI
Models Business3 min
Claude Haiku 5.5 costs 90% less, until prompts get long
Anthropic released Claude Haiku 5.5 at $0.10 per million input tokens and $0.50 output for prompts up to 100,000 tokens, a tenth of Haiku 4.5's price. Above that line the deal changes, so check your prompt lengths before you celebrate.
Why it matters If your calls are short (routing, extraction, sub-agents), your small-model bill can drop sharply this week. If you feed it long documents, the price is five times higher past 100,000 tokens and GPT-6 Luna may be cheaper. Agents on Sonnet 5.5 also get cache reads at half price.
SRCAnthropic
Safety Policy3 min
Common Sense rates ChatGPT for Teens unacceptable
Common Sense Media rated ChatGPT for Teens "Unacceptable Risk" after more than 4,000 test prompts, finding parent alerts rarely fired and hotline referrals fell after launch. OpenAI disputes the timing, but the referral numbers are the part its answer doesn't cover.
Why it matters If you are a parent, don't treat ChatGPT's parent alerts as supervision: in these tests they mostly stayed silent. If you build AI products teens use, this report is the test plan the next reviewer will run on you: hotline naming, alert delay and easy bypasses.
SRCCommon Sense Media
Agents Hardware3 min
Microsoft fences in Windows agents as Nvidia PCs ship
Microsoft made its agent containment layer generally available on Windows 11 and opened preorders for Nvidia RTX Spark PCs, from $2,599.99 for the Surface Laptop Ultra. The laptop gets the headlines, but the fence around your agents is the bigger change.
Why it matters If you run IT, you can now write a policy for which files and networks an AI agent may touch on a Windows 11 PC, enforced by the operating system. If you were waiting on local hardware for 120B-class coding models, prices and ship dates are finally set.
SRCMicrosoft
Safety Creative2 min
Google opens SynthID Detector to everyone, blind spots included
Google opened SynthID Detector worldwide, letting anyone check images, video and audio for watermarks from Google, OpenAI, Nvidia and Kakao. A yes from it means something. A no means almost nothing.
Why it matters If you verify media for a living, you now have a free first check for content made with Google or OpenAI tools. Just never read a clean result as proof a file is real: content from models without SynthID passes straight through.
SRCGoogle

ShareX LinkedIn Hacker News Email

Prompt of the day

Find out what this week's small-model price cuts do to your bill

› I run [describe your app, agent or workflow] on [current model and its price per million input and output tokens]. A typical month looks like this: [number of requests], an average prompt of [tokens], an average reply of [tokens], and [share]% of prompts go over 100,000 tokens.

Two small models now list the same base price of $0.10 per million input tokens and $0.50 per million output tokens. Option A rises to $0.50 input and $2.50 output once a prompt passes 100,000 tokens. Option B rises to $0.20 input and $0.75 output once a prompt passes 272,000 tokens. Assume Option A may use up to 25% more tokens for the same text because of a newer tokenizer.

1. Calculate my monthly cost on my current model, Option A and Option B. Show the math in a table.
2. Tell me the prompt length at which the cheaper option flips.
3. Suggest three changes to my prompts, caching or routing that would cut the bill further, with the rough saving for each.

Before you start, ask me for any number above I left blank. Do not guess my usage.

Open in Magai

Works in any AI chat. In Magai, run it against several models at once.

Get every edition

This one arrived in inboxes at 6am ET.

OpenAI brings GPT-6 to every ChatGPT plan, with clickable answers

Claude Haiku 5.5 costs 90% less, until prompts get long

Common Sense rates ChatGPT for Teens unacceptable

Microsoft fences in Windows agents as Nvidia PCs ship

Google opens SynthID Detector to everyone, blind spots included