Dots, Sol and Space: What OpenAI Shipped at DevDay 2026
OpenAI priced GPT-6.1 Sol at $2 per million input tokens and $10 per million output tokens. That is one-fifth of the $10 and $50 it charges for GPT-6 Astra. The company reports that the cheaper model matches Astra on DeepSWE v1.1, a benchmark of long-horizon software engineering tasks. The figure appeared on September 29 in OpenAI's DevDay 2026 recap, which lists more than 20 announcements across ChatGPT, Codex, the API and enterprise tooling. OpenAI also said the ChatGPT platform now reaches a collective 1.2 billion weekly users. Availability differs sharply from item to item, so the plan tier and release status matter as much as the headline.
Dots
The lead launch is dots, agents that run continuously on GPT-6 Astra with their own cloud computer and browser. OpenAI says they can reach more than 4,000 apps through plugins. Users can message or call a dot in ChatGPT, Slack or Teams, with texting listed as coming. The user can open the dot's computer to inspect its work and can optionally let it connect to a laptop. Dots are rolling out on Pro and Business Premium plans in eligible markets. Enterprise, Edu and Healthcare workspaces get a beta that stays off until an administrator enables it. The first dot is included in the plan price, and conversations with it do not count against usage limits. Tasks it starts in Codex or ChatGPT Work do count.
OpenAI describes several controls. When a dot is idle it does "proactive research" using only read-only connections to the user's apps, so it cannot send messages or change content in that mode. Custom Rules let a user allow, require approval for, or block specific actions, and an auto-review step checks actions that could affect accounts or share information. Changing a password always stays with the user. OpenAI's published example of a dot at work involves an early tester whose dot noticed a forgotten invoice, prepared it, and sent it after he approved. The company also says its own engineers use dots to investigate bugs in Slack. OpenAI acknowledges that dots can still make mistakes and advises reviewing consequential work. The BenchLM write-up of the event points out that OpenAI supplied these examples without a measured success rate across customer projects, and no such rate appears in the announcement.
Specialist dots are a separate track. They receive their own identity, credentials and IT-provisioned hardware, and they are limited to focused enterprise pilots. OpenAI names procurement, invoice processing, email marketing, customer support and commercial contracting as the areas tested internally. It is also working with Microsoft to bring them under Agent 365 governance controls.
A cheaper Sol and a faster Astra
GPT-6.1 Sol is available to Plus, Pro, Business, Enterprise and Edu users in ChatGPT Work and Codex, and through the API as gpt-6.1-sol. It is not yet in the standard Chat experience. Cached input costs $0.10 per million tokens. On DeepSWE v1.1, OpenAI says Sol matches Astra at roughly one-fifth of the cost and beats GPT-6 Sol's best score by 6.4 percentage points. On OSWorld 2.0's offline set it lands within 2.1 points of Astra at roughly one-seventh of the cost per task. On AutomationBench it scores 2.2 points above Opus 5.5 at medium reasoning effort, at about a third of the cost.
The science result carries a caveat that OpenAI states itself. On Terminal-Bench Science 0.1, Sol costs $5.47 per task at maximum effort, against $23.21 for Opus 5.5 and $23.80 for Astra. Astra still posts the highest score at 68.1 percent, and OpenAI recommends it for the hardest research tasks. On factuality, Sol cuts the share of responses containing an error from 11.4 to 7.7 percent, but the test uses conversations where users had flagged an earlier model's mistake, and OpenAI says those prompts are not representative of typical use.
Outside numbers are thin so far. A summary in Latent Space's AINews reports that Artificial Analysis places Sol one point below Astra on its Intelligence Index, at $0.72 per task against $3.26, and that Sol uses 10 to 30 percent more output tokens than GPT-6 Sol. Business Standard reports that in OpenAI's cybersecurity testing Sol showed higher exploit-generation success than the earlier Sol model while staying below Astra. One footnote in OpenAI's own post says the cost shown for Claude Fable 5.1 on AutomationBench omits the price of fallbacks, which occurred on about 40 percent of tasks, so comparisons across vendors depend on how those costs are counted.
Speed has its own tier. Ultrafast offers up to 8 times faster generation in Codex, about 300 tokens per second, and up to 6 times in the API. GPT-6 Astra Ultrafast is available now on the new Pro 500 plan, which carries 25 times the Plus allowance, and on Enterprise. Sol Ultrafast is listed as coming soon.
Codex, the API and enterprise data controls
Codex can now run in the cloud, from a phone, or on a computer, with reusable environments that give a team a shared setup and approved permissions. The CLI accepts voice input and adds an /agents view for delegating and tracking several tasks. A new code review experience in the desktop app summarizes diffs and can take an automatic first pass in the cloud for GitHub pull requests and GitLab merge requests. Codex Security Cloud scans whole repositories on demand or on a schedule, investigates findings, removes duplicates and prepares fixes while a laptop is closed. It includes access to models offered through Daybreak Blue.
On the API side, the Agents API now supports computer use and brings in Codex's multi-agent features, tool search and context compaction, with OpenAI running the infrastructure. The recap links to an earlier introduction of that API, and a secondary report dates its first announcement to September 10, so DevDay restated it rather than debuting it. The Decisions API is in limited preview and directs the Luna model at developer-defined questions with a fixed set of answers, for tasks such as routing requests or choosing an agent's next action. Bedrock Managed Agents, built with Amazon, lets OpenAI agents run entirely inside AWS.
For data-sensitive customers, Private Intelligence adds Zero Data Retention with Private Safety Processing, which allows automated safety review without OpenAI staff seeing the content. A Private Inference preview combining confidential computing is listed for this fall.
ChatGPT as a shared workspace
ChatGPT Space gives a team a dedicated area where teammates, ChatGPT and each person's dot work from shared material. It is available on Pro, Business and Enterprise plans on desktop and web, with mobile creation and editing still to come. Pages are documents that people and agents edit together, and collaborative slides, promised within weeks, will export to PowerPoint or Google Slides. Business and Enterprise plans get team tasks that run on a schedule or in response to events such as a new email or Slack message. They also get @ChatGPT in Slack and Microsoft Teams, which works without an individual license for each teammate. A Meetings plugin, in beta on macOS, writes notes into Space and deletes the audio once the notes are ready.
Developers get plugin extensions, which place a plugin in the ChatGPT sidebar with interactive panels and custom file viewers. There is also a Plugin Creator, a redesigned submission flow, and support for the proposed MCP Events specification so plugins can start automations when something changes in a connected app. Sites can now host supported plugins for teammates who use their own connected data and permissions.
Two commercial announcements complete the list. Sign in with ChatGPT lets Plus and Pro users spend their plan allowance across 16 partner tools, including Cognition's Devin, Notion and Vercel. The OpenAI Marketplace lets eligible enterprise customers apply part of an existing OpenAI commitment toward partner software from a first group of 32 vendors, including Figma, Salesforce, ServiceNow, Harvey, CrowdStrike and Baseten.
Author bio
David Borish is the author of The Tony Hawk Paradox: When Video Games Predict Reality and writes about AI at davidborish.com.
