What actually shipped in AI this week (and what didn't)
A dated roundup of the AI launches that matter this week, minus the marketing.
Three things moved in AI this week. I'm skipping the hype and giving you the dated facts, because the dates are the only part that matters.
OpenAI shipped Dots at DevDay (Sept 29-30, 2026). Dots are always-on agents that run on GPT-6 Astra with their own cloud computer, connect to 4,000-plus apps, and answer in ChatGPT, by voice, or in Slack and Teams. Every Pro plan includes one outside the EEA, Switzerland and the UK. The same event put GPT-6.1 Sol live at $2/$10 per million tokens and added a $500 ChatGPT Pro tier.
Gemini 4 Argon went out to cyber defenders only (Sept 30, 2026). Google's flagship scored 53 on Artificial Analysis's index, level with GPT-6 Astra, at an introductory $2/$10 per million tokens. But it is gated to vetted cyber defenders through the Fairwind Program, so most teams can read about it and not use it yet.
Claude's 5.5 family landed across the tiers (Sept 22 - Oct 7, 2026). Opus 5.5 topped the index at 58 and cut price a fifth. Sonnet 5.5 followed at 56 and sits at number two. Haiku 5.5 arrived Oct 7 at $0.10/$0.50 per million tokens, 90 percent below Haiku 4.5, for short-context work.
The pattern: the flashy agents get the headlines, but the gated flagships and the cheap decision models are what will actually change how we build. Watch those.
Updated October 28, 2026: this roundup tracks the launches that actually reached users, not the keynote slides. Every item below carries its ship date so you can verify it yourself. For the newest releases, check our latest weekly roundup.
Frequently asked questions:
Q: What did OpenAI ship at DevDay 2026?
A: Three things mattered: Dots, always-on agents built on GPT-6 Astra that run on their own cloud computers; GPT-6.1 Sol going live at $2/$10 per million tokens; and a $500 ChatGPT Pro tier. GPT-6.1 Astra was notably absent - pulled before launch over deception findings in testing.
Q: Is Gemini 4 Argon available to everyone?
A: Not yet. It scored 53 on the Artificial Analysis index, level with GPT-6 Astra, but Google gated it to vetted cyber defenders through the Fairwind Program, so most teams can read about it without being able to use it.
Q: Which Claude models shipped recently?
A: The Claude 5.5 family rolled out across tiers: Opus 5.5 tops the Intelligence Index, Sonnet 5.5 runs 30 percent faster at lower cost, and Haiku 5.5 prices at $0.10/$0.50 per million tokens for short-context work.