OpenAI DevDay 2026 Roundup: Dots, GPT-6.1 Sol Pricing, Ultrafast

OpenAI published its DevDay 2026 Recap on September 29, 2026, listing more than 20 product and API updates across ChatGPT, Codex, models, agents, plugins, and enterprise security. The company called the event its biggest DevDay yet. That is OpenAI’s framing, not an independent ranking.

This roundup sticks to OpenAI’s own posts, docs, and help articles for hard facts. Benchmarks, speed multipliers, and “near-Astra” language are labeled as OpenAI’s claims. Unannounced items are labeled as such.

One naming trap matters upfront. Our earlier coverage of OpenAI cancelling GPT-6.1 Astra still stands for that specific model name. DevDay materials ship GPT-6 Astra (including Ultrafast) and introduce GPT-6.1 Sol. Do not treat those as the same release.

What shipped on the keynote stage

OpenAI grouped the day into new ways of working, Codex and API tooling, ChatGPT plugins, collaboration features, and subscription changes. The headline consumer product is Dots: always-on agents OpenAI says are powered by GPT-6 Astra, with their own cloud computer and plugin access to over 4,000 apps.

On the model side, GPT-6.1 Sol is the priced upgrade path OpenAI pushes for coding, computer use, and professional workflows. Ultrafast is the paid speed tier for GPT-6 Astra now, with GPT-6.1 Sol Ultrafast listed as coming soon.

For builders, DevDay pushed computer use into the Agents API, Codex cloud and CLI updates, a Decisions API limited preview on Luna, Zero Data Retention with Private Safety Processing, and a deeper plugin surface inside ChatGPT.

Announcement table: availability and USD prices

Prices below are USD and come from OpenAI’s pricing docs or help pages where a dollar figure is published. Cells marked “not listed” mean OpenAI did not publish a standalone dollar price for that row in the materials checked for this article.

Announcement	What it is	Availability (per OpenAI)	Price (USD)
Dots	Always-on agents with cloud computer; OpenAI says powered by GPT-6 Astra	Pro and Business Premium in eligible markets; Enterprise/Edu/Healthcare beta when admin enables (off by default). Pro personal rollout excludes EEA, Switzerland, and the UK	Included with eligible plans; OpenAI says dots usage will not count toward plan allowances for the next month, then will publish usage terms
GPT-6.1 Sol (API gpt-6.1-sol)	Upgrade to GPT-6 Sol for agentic coding, computer use, professional work	API today; ChatGPT Work and Codex for Plus, Pro, Business, Enterprise, Edu. Not yet in Chat. Ultrafast coming soon	Standard short-context API: $2 / $0.10 cached / $2.50 cache write / $10 output per 1M tokens (OpenAI pricing table)
GPT-6 Astra standard API	Frontier model used across Codex, ChatGPT Work, Agents API examples	API and product surfaces listed in DevDay materials	Standard short-context: $10 / $1 cached / $12.50 cache write / $50 output per 1M tokens
GPT-6 Astra Ultrafast	Paid speed tier via service_tier: ultrafast	API now (low default rate limits); ChatGPT Work and Codex on Pro 500 and Enterprise. OpenAI claims up to 8x Codex / 6x API speed (~300 tok/s)	Ultrafast short-context: $60 / $6 cached / $75 cache write / $300 output per 1M tokens; long-context: $120 / $12 / $150 / $450
Pro 500	New ChatGPT Pro tier with highest usage and Ultrafast	Available now	$500 per month; OpenAI says 25x ChatGPT Plus allowance
Pro 100 / Pro 200	Other Pro tiers	Available; Pro 200 open to new subscriptions again	$100 / $200 per month; Ultrafast not included
Agents API + computer use	Managed agent harness; browser/computer-use tool in OpenAI-hosted env	API; also Codex and ChatGPT Work on Pro 500 and Enterprise for computer use	Billed via underlying model/API usage; no separate flat fee listed in the recap
Decisions API	Luna-focused classification over finite answer sets	Limited preview today; broader release planned in coming days	Not listed as a standalone product price
ZDR + Private Safety Processing	Automated safety review without OpenAI personnel reading customer content	Documented for eligible API customers; Private Inference preview “this fall”	Contact / enterprise terms; no public list price in the recap
Codex cloud / CLI / Code Review / Security Cloud	Remote Codex, voice CLI, desktop review, repo security scanning	Varies: cloud on Plus through Enterprise; CLI and Code Review all plans; Security Cloud for Pro, Business, Enterprise, Edu	Included with eligible ChatGPT/Codex plans; Daybreak Blue models referenced for Security Cloud
Plugin extensions, MCP events, Sites hosting plugins	Sidebar UIs, event-driven automations, plugins inside Sites	Extensions and MCP events: all plans; Sites hosting plugins: Business, Enterprise, Healthcare, Edu	Not listed separately
ChatGPT Space, Pages, collaborative slides	Team workspace, collaborative docs, slides (slides soon)	Space/Pages: Pro, Business, Enterprise; slides in coming weeks for same plans	Included with those plans
Meetings plugin	Local note-taking into ChatGPT Space; audio deleted after notes	Beta on macOS desktop for Pro and Business; Enterprise coming soon	Included with those plans
Sign in with ChatGPT / plan usage in partner apps	Identity plus spending ChatGPT allowance in partner tools	Identity global; plan usage for Plus and Pro in participating tools	Uses existing plan allowance / credits; partner list includes Devin, Notion, Vercel, and others
OpenAI Marketplace	Apply OpenAI commitment toward approved partner software	Eligible enterprise customers can express interest; 32 launch partners named	Uses existing enterprise commitment; partner SKUs vary

Dots: always-on agents with a hard regional cut

OpenAI describes Dots as proactive assistants that keep working across projects, learn preferences over time, and can reach you in ChatGPT, Slack, or Teams. Official copy says each Dot has its own cloud computer and browser, and can connect to other devices when you allow it.

Availability is plan-gated and geography-gated. Per OpenAI help, Dots roll out on web, mobile, and desktop to Pro users in markets excluding the European Economic Area, Switzerland, and the UK. Business Premium users get Dots across all supported ChatGPT regions. Enterprise (and related workspace) users need an admin to turn the beta on.

For readers in the United States, Canada, Australia, and India, personal Pro Dots sit outside the EEA/UK/Switzerland exclusion as written today. Business Premium availability follows supported ChatGPT regions. Rollout is gradual; re-check OpenAI’s eligible-markets help before buying for Dots alone.

OpenAI also says that for the next month, Dots usage will not count toward eligible Pro, Business, and Enterprise allowances, with usage terms to follow. Treat that as a temporary vendor policy, not a permanent free tier.

GPT-6.1 Sol: the priced “near-Astra” API play

OpenAI’s Sol post positions GPT-6.1 Sol as nearly matching GPT-6 Astra on agentic coding, computer use, and professional work at about one-fifth of Astra’s standard input and output token prices. Those performance comparisons are OpenAI’s own evaluations on DeepSWE, GDP.pdf, AutomationBench, OSWorld 2.0, and related tests.

Confirmed commercial facts from OpenAI:

• API model id: gpt-6.1-sol
• Standard list prices published with the launch: $2 input / $0.10 cached input / $10 output per million tokens (OpenAI’s announcement text). The public pricing table also shows a $2.50 cache-write column for short context.
• Available in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise, and Edu
• Not yet available in Chat
• GPT-6.1 Sol Ultrafast “in the coming days,” with OpenAI claiming up to 8x faster token generation versus standard Sol in Codex

If you already tracked the September 22 cut on GPT-6 Sol and Luna pricing, treat 6.1 Sol as the capability bump on the same $2/$10 list lane, with a cheaper cached-input rate than the earlier Sol cached price OpenAI cites.

GPT-6 Astra Ultrafast and Pro 500

Ultrafast is documented as the fastest API service tier. Docs say it is broadly available for GPT-6 Astra, with preview-style language still attached historically to older Sol variants. For DevDay, the recap is clear: GPT-6 Astra Ultrafast is available today in the API and in ChatGPT Work and Codex on Pro 500 and Enterprise. GPT-6.1 Sol Ultrafast is coming soon.

OpenAI claims up to 8x faster token generation in Codex (about 300 tokens per second) and up to 6x in the API. Those are vendor speed claims. Default Ultrafast rate limits on Astra are low: OpenAI docs list 500,000 TPM for tiers 1-3, 1,000,000 TPM for tier 4, and 5,000,000 TPM for tier 5. Higher limits need an account team request.

ChatGPT Pro tiers now include Pro 100 ($100/mo), Pro 200 ($200/mo), and Pro 500 ($500/mo). Only Pro 500 includes Ultrafast among Pro plans. OpenAI says Pro 500 carries 25 times the ChatGPT Plus allowance. Pro 200 grandfathering of older allowances runs through October 29, 2026 and does not add Ultrafast.

Agents API, computer use, Decisions, and Bedrock

The Agents API launched earlier in September as a public beta that exposes the Codex-style harness: sessions, tools, multi-agent fan-out, and hosted or partner sandboxes. DevDay’s addition is computer use: agents navigate a browser in an OpenAI-hosted environment, with your app approving site access and sign-in steps.

Decisions API focuses Luna on finite answer sets for classification and routing: limited preview now, broader release planned soon.

OpenAI also announced Bedrock Managed Agents with Amazon so agents can run in AWS; see AWS/OpenAI docs for details.

Private Intelligence and ZDR with Private Safety Processing

Zero Data Retention with Private Safety Processing is OpenAI’s answer for customers who want automated safety review without OpenAI staff reading prompts and responses. Customer content lands in customer-controlled storage under customer key control; review runs in a hardware-attested runtime that OpenAI says disables human access.

Private Inference is still a preview “coming this fall.”

Codex: cloud, CLI, review, and security scanning

Codex in the cloud can run from a computer, phone, or any device, with reusable environments for shared team settings. OpenAI lists Plus, Pro, Business, Healthcare, Education, and Enterprise. The CLI adds voice control and an /agents view; Code Review and Security Cloud cover PR review and repo scanning on the plan tiers listed in the table above.

ChatGPT product surface: plugins, Space, Meetings, identity

ChatGPT Space is a team home for shared knowledge with your Dot. Pages are collaborative documents. Collaborative slides are scheduled in the coming weeks. Business and Enterprise get team sharing and @ChatGPT in Slack and Microsoft Teams without every teammate needing an individual license for that mention flow.

The Meetings plugin (macOS desktop beta for Pro and Business) takes notes without a bot joining the call; OpenAI says audio is deleted once notes are ready. Shareable profiles collect Sites and plugins for reuse across Free through Enterprise plans.

Sign in with ChatGPT spreads identity globally and lets Plus and Pro users spend plan allowance across listed partner apps, with per-app controls. OpenAI names 16 partners including Devin, Notion, and Vercel.

What it means for Indian developers

API pricing is USD-first worldwide, so GPT-6.1 Sol at $2/$10 per million tokens and Astra Ultrafast at $60/$300 short-context list rates hit Indian startups the same way they hit US ones: Ultrafast is a specialist lane, not a default. For day-to-day agent loops, Sol’s cached input at $0.10 per million tokens is the lever OpenAI is advertising.

Dots on personal Pro are not blocked by the EEA/UK/Switzerland exclusion language, so Indian Pro subscribers are in the eligible-markets story OpenAI published, subject to gradual rollout. Business Premium Dots follow supported ChatGPT regions. If your company is on a standard Business plan rather than Business Premium, check the plan name carefully before promising Dots to a client.

Computer use and Agents API matter for Indian product teams building browser workflows without standing up their own agent runtime. Rate limits on Ultrafast start low, so production traffic still needs tier upgrades or an account conversation.

OpenAI says DevDay Exchanges continue later this year in Bengaluru, Tokyo, Seoul, Paris, Berlin, London, Sao Paulo, and Mexico City. Sign-ups run through OpenAI’s Exchange interest form. Official DevDay site copy says “later this year” for Bengaluru; it does not publish a confirmed calendar date on that page in the materials checked here, so treat city confirmation as firm and any specific date from secondary posts as unverified until OpenAI lists it.

Separately, the pre-DevDay shutdown of four legacy models still affects older Indian codebases that never left instruct-era IDs. See our note on the September 28 legacy model shutdown if you still have those strings in production.

USA, Canada, and Australia notes

Personal Pro Dots are written as available outside the EEA, Switzerland, and UK exclusions, which covers the US, Canada, and Australia under OpenAI’s current help text. Business Premium Dots follow supported ChatGPT regions in those countries as well.

Pro 500 at $500/mo is a USD list price aimed at heavy Codex and Ultrafast users.

Shipping now versus still previewed

Shipping or broadly listed as available now (with plan gates): Dots (eligible markets), GPT-6.1 Sol in API/Work/Codex, GPT-6 Astra Ultrafast in API plus Pro 500/Enterprise product surfaces, Pro 500, Agents API computer use on listed surfaces, Codex cloud/CLI/Code Review/Security Cloud on listed plans, plugin extensions and MCP events, Space and Pages, shareable profiles, Sign in with ChatGPT identity, Marketplace interest for enterprises.

Preview, alpha, or “coming soon” in OpenAI’s own words: GPT-6.1 Sol Ultrafast; Decisions API limited preview; Private Inference this fall; collaborative slides in the coming weeks; Meetings Enterprise generally available later (alpha with selected customers); specialist Dots preview for IT-provisioned org use; texting beta limited to some US Pro users and not for Business/Enterprise.

Not announced in the official DevDay materials checked here: a public GA date for GPT-6.1 Astra (that model was cancelled before the keynote); a confirmed Bengaluru Exchange calendar date on the DevDay site; a free consumer Dot tier; Ultrafast on Pro 100/Pro 200.

FAQ

Is GPT-6.1 Astra available after DevDay 2026?

No. OpenAI cancelled GPT-6.1 Astra after alignment tests missed its own bar, which we covered separately. DevDay shipped GPT-6 Astra (including Ultrafast) and GPT-6.1 Sol. Keep the version strings straight in your runbooks.

What does GPT-6.1 Sol cost on the API?

OpenAI’s launch post lists $2 per million input tokens, $0.10 per million cached input tokens, and $10 per million output tokens. The public pricing table also shows a $2.50 cache-write rate for short context. Ultrafast pricing for Sol is not listed as live yet.

Can I get Dots on ChatGPT Pro in India, the US, Canada, or Australia?

OpenAI’s help text excludes the EEA, Switzerland, and the UK from the Pro personal rollout. India, the US, Canada, and Australia are not in that exclusion list. Access still rolls out gradually, and Business users need Business Premium or an admin-enabled Enterprise beta.

Who gets Astra Ultrafast in ChatGPT and Codex?

OpenAI says ChatGPT Work and Codex Ultrafast for Astra are on Pro 500 and Enterprise. Among personal Pro plans, only Pro 500 includes Ultrafast. API Ultrafast is available more broadly at low default rate limits.

Where should developers start if they only have a day?

Read the official recap, try gpt-6.1-sol against your current Sol or Sonnet workloads, and only enable Ultrafast where latency pays for a 6x to 30x list-price jump on Astra tokens. If you ship browser agents, start with the Agents API computer-use guide rather than rebuilding a harness.