OpenAI planning $500/mo Pro Max? A ChatGPT tier leaks, TestingCatalog says
A $500 tier with "Fastest Work and Codex" sits in unreleased ChatGPT plan text, TestingCatalog says; OpenAI hasn't confirmed it. Two more stories inside.
A $500-a-month ChatGPT Pro Max plan has surfaced in unreleased subscription references, TestingCatalog reported Thursday, four days before OpenAI's DevDay. The only listed difference from the $200 Pro tier is "Fastest Work and Codex," and OpenAI has said nothing. Also today: Anthropic will bill again for API requests its safety classifiers block before Claude answers, in three categories where it says false positives are low, on every platform Claude runs on. And Whiteboard, an open-source Show HN launch, gives Claude Code and Codex a canvas to draw their plans on.
What changed this week
OpenAI planning $500/mo Pro Max? A ChatGPT tier leaks, TestingCatalog says
A ChatGPT Pro Max plan "will cost $500," TestingCatalog posted Thursday, from "unreleased ChatGPT subscription references" posted earlier by Tibor Blaho. Next to the $200 Pro tier, the only listed difference is "Fastest Work and Codex"; higher usage limits "may also be part of the package, although that is not confirmed yet." The site guesses Cerebras hardware behind the speed and says "OpenAI has not confirmed" it. OpenAI has said nothing, and DevDay is Tuesday. Treat the price as a leak until then.
OPENAI 🔥: The upcoming ChatGPT Pro Max plan will cost $500.
— 🚨 AI News | TestingCatalog (@testingcatalog) September 24, 2026
So far, this will be one of the most expensive AI subscriptions available on the market.
I hope "it will be worth the wait" 👀 https://t.co/62fGntpulu pic.twitter.com/uurB8yIOnr
Anthropic bills 3 refusal types again, even when Claude never answers
Anthropic will again charge for API requests its safety classifiers decline before Claude produces any output, its @ClaudeDevs account posted Thursday, citing "coordinated attacks." The post names the billed categories as "biology, distillation attacks, and frontier LLM development"; the docs never say "distillation" and list bio, frontier_llm and reasoning_extraction, a request that "asks the model to reproduce its internal reasoning." Cyber and general-harms refusals still aren't billed, and the rule applies "on every platform" Claude runs on. Expect those three refusals on the invoice.
Today, we'll resume charging for requests our safeguards block before Claude responds. This only applies in categories with low false positive rates: biology, distillation attacks, and frontier LLM development. We've seen some coordinated attacks on our systems in recent weeks,…
— ClaudeDevs (@ClaudeDevs) September 24, 2026
Whiteboard gives coding agents a shared design canvas
Whiteboard, a Show HN launch from a startup in Y Combinator's winter 2026 batch, is "an open-source desktop app where humans and agents can architect software together in a common workspace," its README says. It "plugs into the tools you already use," naming Claude Code and Codex, and "gives your agent an SDK to draw on an in-app canvas to describe its work." It climbed from 65 points to 333 overnight on Hacker News, mostly reactions, not reports from people who've run it. Free and open source, for macOS and Fedora.
Know someone who'd want this in their inbox? Forward it — that's how this grows. And if we got something wrong, or you think we buried the real story today, hit reply. A person reads every one.
The New Way is written with AI. It gathers the day's stories, checks them against their sources and drafts every summary. A person decides what runs and reviews every issue before we hit send.