Key Takeaways –
-
Dots: OpenAI’s always-on agents can handle goals, use apps, and work through cloud computers with user approval for sensitive actions.
-
GPT-6.1 Sol: The model delivers near-GPT-6 Astra performance for complex tasks at much lower API costs.
-
Agent Tools: Codex Cloud, ChatGPT Space, Agents API, and Ultrafast expand AI support across coding, teamwork, and software workflows.
At its latest DevDay, OpenAI moved artificial intelligence well beyond the conventional chatbot framework by introducing more than 20 major product updates spanning ChatGPT, Codex, underlying models, APIs, and workplace solutions.
The primary focus centers on autonomous AI agents capable of managing tasks, operating software, navigating multiple applications, and executing long-running projects with minimal direct human intervention. According to OpenAI, ChatGPT now serves 1.2 billion weekly users, instantly providing these new features with a massive user base.
Dots Put AI Agents at the Center
Dots emerged as the flagship announcement of DevDay. OpenAI characterizes Dots as continuous, always-on agents driven by GPT-6 Astra. Every Dot operates within its own dedicated cloud computer and browser environment, while also integrating with over 4,000 third-party applications via OpenAI’s plugin network. A Dot can pursue objectives around the clock, maintain continuity across multiple assignments, and adapt through ongoing feedback.
These agents can independently execute workflows without requiring step-by-step human guidance. OpenAI highlighted scenarios such as a software developer routing customer feedback directly into tested bug fixes, a product manager refreshing launch documentation following a scope adjustment, or a researcher integrating fresh data sets into an ongoing analysis. Additionally, a Dot can generate and dispatch invoices pending user authorization. The rollout of Dots has begun for Pro, Business Premium, and Enterprise subscribers in supported markets.
Safety measures are deeply embedded into the architecture of Dots. Users retain the ability to establish operating limits, monitor progression, and manually authorize critical actions. Because each Dot functions inside a private cloud environment, the user’s personal device remains isolated unless explicit permissions are granted. Furthermore, Dots integrate directly with ChatGPT, Slack, and Microsoft Teams, embedding themselves smoothly into existing corporate workflows.
GPT-6.1 Sol Brings Lower AI Costs
GPT-6.1 Sol provides developers with an economical alternative for demanding software engineering, computer navigation, and professional duties. OpenAI states that Sol achieves performance levels comparable to GPT-6 Astra while costing one-fifth of Astra’s standard token input and output pricing. Standard API charges are set at USD 2 per million input tokens, USD 0.10 per million cached input tokens, and USD 10 per million output tokens.
The model performs strongly across standard performance benchmarks. On DeepSWE v1.1, GPT-6.1 Sol matches the output of GPT-6 Astra at roughly 20% of the expense. On OSWorld 2.0, Sol reaches within 2.1 percentage points of Astra’s accuracy under maximum reasoning parameters while consuming approximately one-seventh of the per-task cost. OpenAI also notes that Sol’s factual error rate drops to 7.7% at low reasoning settings, an improvement from the 11.4% error rate recorded by GPT-6 Sol.
Also Read – 10 Best Cloud Security Platforms for Developer-Led Enterprises in 2026
Ultrafast Targets Speed
OpenAI also rolled out Ultrafast, a high-performance speed tier built for GPT-6 Astra. The firm reports token generation rates up to 8 times faster within Codex—reaching speeds as high as 300 tokens per second—along with API output speeds up to 6 times faster. Astra Ultrafast is immediately accessible via the API, ChatGPT Work, and Codex environments on Pro 500 and Enterprise tiers, with GPT-6.1 Sol Ultrafast slated for release soon.
Codex Moves into the Cloud
Codex now equips developers with expanded capabilities to execute software operations independently of local hardware. Codex Cloud executes tasks within cloud environments managed entirely by OpenAI, which house project repositories, essential utilities, dependencies, and security configurations. Developers can initiate or resume workflows seamlessly across desktop, web, and mobile interfaces.
In addition, OpenAI integrated voice command capabilities into the Codex CLI alongside a new Agents view, automated Code Review utilities, and Codex Security Cloud. Security Cloud continuously monitors GitHub repositories, analyzes anomalies, filters redundant alerts, and drafts remediation patches while local workstations remain powered down.
ChatGPT Space Creates a Shared Work Hub
ChatGPT Space unifies Pages, file repositories, project details, team members, and AI agents into a centralized collaborative workspace. Teams can leverage Space to organize initiatives, utilize Pages for dynamic documents and data visualizations, and take advantage of collaborative presentation slides, Team Tasks, and native integration within Slack and Microsoft Teams. The system also features a Meetings plugin capable of automatically compiling transcripts and action items, currently available in beta on macOS for Pro and Business accounts.
New APIs Give Developers More Control
The new Decisions API enables developers to supply Luna with specific query sets paired with predetermined response options. This capability allows systems to categorize text inputs, direct traffic, or determine subsequent agent actions. OpenAI has launched this functionality as a limited preview ahead of a broader public release.
Additionally, the Agents API now supports computer-use functionality. Developers can construct agents capable of interacting directly with software interfaces through an OpenAI-managed browser session. This updated API introduces multi-agent orchestration, tool search functions, dynamic tool execution, and context compaction features, furnishing developers with a robust foundation for building advanced agent frameworks.
Also Read – Agentic AI Market Outlook 2026-2035: Market Size, Growth, Adoption and Key Opportunities
Pro 500 Raises the Top Tier
OpenAI introduced the Pro 500 subscription tier, priced at USD 500 per month. This tier delivers 25 times the usage capacity of the Plus subscription alongside full access to Astra Ultrafast. Simultaneously, the Pro 200 subscription option has been reinstated, though newly registered Pro 200 accounts will operate under distinct usage caps compared to certain legacy accounts.
The underlying theme of DevDay 2026 is unmistakable: OpenAI aims to transform AI from a conversational assistant into an active executor. Products like Dots, Codex Cloud, the Agents API, ChatGPT Space, and GPT-6.1 Sol collectively signal an industry-wide transition toward intelligent software capable of accepting a broad objective, utilizing software tools, and completing complex assignments far in advance of human intervention.
FAQs
1. What is OpenAI DevDay 2026?
OpenAI DevDay 2026 is the company’s developer event that introduced more than 20 major updates across models, agents, APIs, Codex, ChatGPT, and workplace tools.
2. What are Dots?
Dots are always-on AI agents powered by GPT-6 Astra. They can use apps, work through cloud computers, handle tasks, and retain context across projects.
3. What is GPT-6.1 Sol?
GPT-6.1 Sol is a lower-cost model for complex coding, computer use, and professional tasks. OpenAI says it offers near-GPT-6 Astra performance at a much lower price.
4. What does Codex Cloud do?
Codex Cloud lets developers run coding tasks on OpenAI-managed cloud computers. Tasks can continue without a local computer running throughout the process.
5. What is ChatGPT Space?
ChatGPT Space creates a shared work area for people, files, projects, Pages, and AI agents, with support for tools such as Slack and Microsoft Teams.
