What OpenAI DevDay 2026 announcements covered
OpenAI held its DevDay 2026 keynote on September 29, 2026, in San Francisco. The flood of OpenAI DevDay 2026 announcements landed fast: new models, a new agent product, new subscription tiers, and a redesigned way teams collaborate inside ChatGPT. By the time the keynote ended, someone on your leadership team had likely already asked which of it you're turning on first.
During the keynote, OpenAI said ChatGPT now has 1.2 billion weekly users, one of the standout figures in the OpenAI DevDay news coverage that day. That figure comes from OpenAI's own reporting and hasn't been independently audited, so treat it as context, not a verified benchmark.
Common practice after OpenAI DevDay 2026 announcements like this is to route the list straight to engineering and wait for documentation to catch up. That gets you inventory, not adoption: knowing a feature exists doesn't tell you which teams will actually use it, or use it well. This rundown of the OpenAI Developer Day announcements groups every enterprise-relevant item by what it changes for you.
Sources: OpenAI DevDay 2026: The biggest news and announcements | The Verge
New models and pricing tiers reshape the cost math

OpenAI's headline new model is GPT-6.1 Sol, priced at a fifth of GPT-6 Astra's standard input and output token rates. OpenAI's own announcement says Sol 'nearly matches' Astra's intelligence on agentic coding, computer use and professional work, a company performance claim that hasn't been independently benchmarked in what's been reported.
Cached input tokens for Sol run $0.10 per million tokens, per OpenAI's announcement text. A new Ultrafast speed tier processes requests faster, at roughly six times the price of standard usage, according to live coverage of the keynote.
OpenAI also added a new subscription tier called Pro 500, priced at $500 a month, and changed the usage multiplier on the existing $200 Pro plan for new subscribers, dropping it from twenty times Plus usage to ten times, per comments from an OpenAI product lead relayed by a secondary outlet. Existing $200 subscribers reportedly keep the old multiplier for a period before it changes.
| Tier | What it is | Price signal | Who's affected |
|---|---|---|---|
| GPT-6.1 Sol | New flagship-adjacent model | About a fifth of Astra's standard token price; cached input at $0.10 per million tokens | Teams choosing between Sol and Astra for agentic and coding work |
| Ultrafast | Faster processing speed tier | Roughly six times standard usage price | High-volume or latency-sensitive workloads |
| Pro 500 | New top subscription tier | $500 per month | Power users who outgrow the $200 Pro plan |
| $200 Pro plan (updated) | Existing top plan, new usage multiplier | 10x Plus usage for new subscribers, down from 20x | Current Pro subscribers, temporarily grandfathered |
Sources: DevDay 2026 Recap | OpenAI, OpenAI makes 20+ announcements at DevDay event including always-on agents called dots, watch live - 9to5Mac, OpenAI DevDay 2026 live blog, OpenAI DevDay 2026: Live updates and announcements, OpenAI DevDay 2026: The biggest news and announcements | The Verge, Everything OpenAI Announced At DevDay 2026, Including Its Muse Alternative
Dots: an always-on agent that raises governance questions

Dots is OpenAI's new always-on agent product. CEO Sam Altman described Dots on stage as 'remarkably capable, always-on agents built to handle really anything you can think of,' language worth reading as marketing from OpenAI's own CEO rather than an independent capability review.
Access rolls out by plan. OpenAI's recap lists Dots as available to Pro and Business Premium subscribers in eligible markets, while one attendee at the keynote reported Pro and Enterprise customers getting access the same day. If your organization spans multiple ChatGPT plans, don't assume everyone has it yet.
The same open-sourced Codex harness that powers OpenAI's coding tool now also runs Dots, according to independent reporting. That's worth flagging to your security team: an agent built on your coding tool's own automation engine inherits its permission model, for better and worse.
- Confirm which plan tiers in your organization actually include Dots before promising broad access.
- Review data and app-connection permissions before letting an always-on agent run unattended.
- Ask whether the Codex harness underneath Dots needs its own security review, since it powers both tools.
Sources: DevDay 2026 Recap | OpenAI, OpenAI DevDay 2026 live blog, OpenAI DevDay 2026: Live updates and announcements, OpenAI Gave AI Agents Their Own Computers at DevDay 2026. Here's Everything It Announced - Decrypt
Codex and the Decisions API extend what agents can do
The new Decisions API points Luna, OpenAI's cheapest model, at a fixed list of possible answers for fast classification tasks, according to independent reporting from the keynote. It was in limited preview at announcement, so its behavior at general availability may look different once it ships broadly.
For a CIO, the practical question isn't whether the API works in a demo. It's whether your team has a concrete classification task worth routing through it, and whether that's worth building against before the preview graduates to general release.
Sources: OpenAI Gave AI Agents Their Own Computers at DevDay 2026. Here's Everything It Announced - Decrypt
Collaboration surfaces and enterprise privacy signals
ChatGPT Space is a new collaborative area described in live coverage as looking like a projects hub where you and your colleagues work alongside your Dots. That's an early impression from the keynote demo, not a reviewed or documented feature.
ChatGPT can also now surface plugin recommendations during a conversation, per the same live coverage. For admins, that means employees may discover and connect third-party plugins mid-chat, which is worth folding into your app-approval process rather than treating as a separate review track.
On the privacy side, OpenAI's Private Intelligence offering is built around zero data retention, according to independent reporting. Full technical detail on how that's implemented wasn't included in what OpenAI shared at the keynote.
Sources: OpenAI Dev Day 2026: Live Updates On The Latest ChatGPT And Codex Announcements, OpenAI Gave AI Agents Their Own Computers at DevDay 2026. Here's Everything It Announced - Decrypt
What remains unsettled, and what to decide next

Several pieces of the OpenAI DevDay 2026 announcements aren't fully live yet. Broad release of the Decisions API, an Ultrafast tier for Sol, and Private Intelligence were all described as coming soon or in the coming days rather than shipped and stable.
We'd treat anything still in limited preview as unsuitable for production planning until OpenAI confirms general availability. Verify the reported $200-plan multiplier change against your own billing before you act on it.
Announcing a feature and a team using it well are two different things. The checklist below is where we'd start.
Sources: OpenAI Gave AI Agents Their Own Computers at DevDay 2026. Here's Everything It Announced - Decrypt
Frequently asked questions
Do we need to buy a new plan to get Dots or GPT-6.1 Sol?
Not necessarily. OpenAI's own recap lists Dots as available to Pro and Business Premium subscribers in eligible markets, so check your existing plan tier before assuming you need to upgrade or buy anything new.
Is the $200 Pro plan change already in effect?
It's not confirmed in OpenAI's own DevDay recap. Verify the reported multiplier change against your own billing and usage dashboard before adjusting budget plans, since terms can shift before OpenAI publishes official documentation.
Should we roll out Dots to every team right away?
We'd start with a single pilot team rather than a blanket rollout. Watch how that team actually uses Dots for a few weeks, then use what you learn to decide whether the access, training and governance work needed to expand it further is worth doing now.
How do we know if employees are actually using any of this?
Access and login counts won't tell you. A structured employee survey covering usage, skills and perceived barriers gives you a clearer read on adoption than assuming a feature is used just because it's turned on.
Sources: DevDay 2026 Recap | OpenAI, Everything OpenAI Announced At DevDay 2026, Including Its Muse Alternative




