OpenAI DevDay 2026: Dots Agents, GPT-6.1 Sol, Codex and 20+ Launches

OpenAI staged its biggest DevDay yet on September 29, packing more than 20 announcements into a single keynote: always-on agents called dots, a new GPT-6.1 Sol model priced at one-fifth of GPT-6 Astra, a premium Ultrafast speed tier, a $500-per-month Pro 500 plan, a major Codex expansion, plugin extensions that turn ChatGPT into an app platform, and an enterprise Marketplace. The company framed the day around one number: ChatGPT now reaches 1.2 billion weekly users, and OpenAI wants developers building directly inside that surface.

This piece walks through every launch that matters for engineers and technical buyers, with pricing and availability for each. Standalone deep-dives on the individual announcements are linked in the cards below; this is the map of the whole day.

The headline act: dots, agents that keep working

The single biggest reveal was dots: persistent, always-on agents that live inside ChatGPT, run on their own cloud computer with their own browser, and keep working between conversations. Powered by GPT-6 Astra, a dot can connect to more than 4,000 apps through the plugin ecosystem, take on multiple projects at once, and message you with progress in ChatGPT, Slack, or Microsoft Teams. OpenAI says dots also do read-only "proactive research" in the background, looking for ways to help before you ask.

Dots are rolling out now to Pro and Business Premium subscribers in eligible markets, with the first dot included at no extra cost; Enterprise, Edu, and Healthcare workspaces get a beta that admins must switch on. OpenAI also previewed specialist dots for organizations, with their own credentials and an integration path into Microsoft Agent 365. Conversations with a dot do not count against ChatGPT usage limits, though Codex and ChatGPT Work tasks a dot kicks off still do.

Models and pricing: GPT-6.1 Sol, Ultrafast, Pro 500

On the model side, GPT-6.1 Sol is the workhorse story: OpenAI says it nearly matches GPT-6 Astra on agentic coding, computer use, and professional work at one-fifth of Astra's token prices — $2 per million input tokens, $0.10 cached, and $10 output, with the API id gpt-6.1-sol. On DeepSWE v1.1 it matches Astra's top-end accuracy while cutting task execution cost by roughly 80 percent. It is live today in ChatGPT Work and Codex for paid plans, but not yet in regular Chat.

For teams that care about latency, the new Ultrafast tier pushes GPT-6 Astra to roughly 300 tokens per second in Codex — up to 8x its standard speed — and up to 6x throughput in the API at six times the standard price: $60 per million input and $300 per million output. Ultrafast anchors the new Pro 500 plan, a $500-per-month subscription with about 25 times the usage allowance of Plus. Less prominently, the $200 Pro plan got trimmed for new subscribers: ChatGPT Work and Codex usage drops from 20x to 10x Plus, and weekly GPT-6 Pro messages fall from 200 to 100.

Codex: cloud environments, voice CLI, review, security

Codex picked up its largest update since launch. It now runs on your computer, remotely from your phone, or entirely in the cloud, with reusable development environments that give teams a shared, permission-approved setup. The CLI was revamped with voice-controlled task steering, an /agents view for delegating several tasks at once, session resuming, and git worktree integration — and OpenAI open-sourced the Codex harness on GitHub.

Two new surfaces sit on top: a dedicated code-review view in the ChatGPT desktop app that summarizes diffs across projects and can run automatic background reviews before GitHub or GitLab feedback goes out, and Codex Security Cloud, which scans repositories on demand or on a schedule, monitors new commits, deduplicates findings, and prepares fixes remotely — with access to Daybreak Blue models included for Pro, Business, Enterprise, and Edu.

The platform play: extensions, Marketplace, identity

The strategic thread of the day was OpenAI opening ChatGPT itself as a development target. Plugin extensions let developers build what OpenAI calls entire applications native to ChatGPT and Codex — a home in the sidebar, interactive panels beside the conversation, and viewers for your product's file types — distributed through OpenAI to all plans. A conversational Plugin Creator, streamlined submission, and in-conversation plugin recommendations round out the developer funnel, and new MCP Events support lets connected apps trigger ChatGPT automations via webhooks.

For the enterprise, the new OpenAI Marketplace lets customers direct existing OpenAI spend commitments toward roughly 32 launch partners, including Adobe, Figma, Salesforce, ServiceNow, Palo Alto Networks, and CrowdStrike. Sign in with ChatGPT went global, and Plus and Pro subscribers can now draw on their plan allowance inside 16 partner tools, including Cognition's Devin, Notion, Vercel, T3, OpenClaw, and Dactyl, with per-app limits.

What developers should verify next

Collaboration features fill out the rest of the list: ChatGPT Space, a shared team workspace on Pro, Business, and Enterprise; collaborative Pages and a Slides preview with PowerPoint and Google Slides export; teams with shared, event-triggered tasks; @ChatGPT mentions in Slack and Microsoft Teams; a Meetings plugin in macOS beta; and shareable profiles. Enterprises also got Private Intelligence — Zero Data Retention with Private Safety Processing now, and a Private Inference preview coming this fall — plus the Decisions API, which returns constrained classifications in about 150 milliseconds using GPT-6 Luna, and computer use landing in the Agents API alongside Bedrock Managed Agents on AWS.

The practical to-do list: check whether your market and plan tier has dots enabled before promising workflows on top of them; benchmark gpt-6.1-sol against your current model, since the cached-input price of $0.10 per million tokens changes agent economics; and if you sell developer tools, read the plugin extension docs before the directory gets crowded. Watch for GPT-6.1 Sol Ultrafast in the coming days, the Private Inference preview this fall, and whether the quiet Pro plan reductions push heavy users toward Pro 500.