PuppyIP Resource Center
AI Tool Updates 8 min read Published 2026-09-30

OpenAI DevDay 2026: GPT-6.1 Sol, Codex, and Agents API at a Glance

At DevDay on September 29, 2026 in US time, OpenAI announced model, Codex, agent API, and ChatGPT extension updates. Development teams should distinguish available features, limited previews, future releases, and the accounts and entry points each supports. This guide follows the official recap.

OpenAI DevDay 2026 GPT-6.1 Sol Codex Agents API Decisions API Dots

Service eligibility and regional restrictions

PuppyIP serves only compliant overseas businesses and their authorized personnel. Proxy services are not available in mainland China. The service may only be used for lawful business activities outside mainland China. Use of this service within mainland China is prohibited.

Hosting a proxy IP or server overseas does not change these restrictions. The service must not be provided to end users in mainland China through relaying, forwarding, sharing or resale. Before use, read the Terms of Service.

Key Takeaways

  • The recap lists more than 20 announcements. This guide focuses on developer choices and integrations without treating every announcement as universally available that day.
  • GPT-6.1 Sol is available through the API and ChatGPT Work and Codex for Plus, Pro, Business, Enterprise, and Edu. Ordinary Chat conversations do not yet have access.
  • Codex adds reusable cloud development environments and updates CLI, code review, and Security Cloud. These are different capabilities with different access.
  • Decisions API was a limited preview that day. Agents API announced computer use, but API access and product-plan access must be judged separately.
  • Dots depends on plans, markets, and administrator settings. Verify plugin extensions, MCP Events, and Bedrock Managed Agents separately rather than treating them as one agent product.

Start with the product entry point

OpenAI scheduled DevDay 2026 for September 29 in US time. Its recap describes more than 20 major announcements across ChatGPT, Codex, models, and developer APIs. Beijing time was already September 30; this is a time-zone difference, not two launches. This page uses verifiable official recap information as of September 30, without merging demonstrations, previews, and general availability.

List your team's entry point first: direct API calls, Codex work, ChatGPT workspace administration, or building agents on AWS. Then inspect that entry point's eligibility, documentation, and console. Appearance in a conference recap does not mean your account, region, or workspace already has access.

GPT-6.1 Sol: Check access before migrating

OpenAI describes GPT-6.1 Sol as an upgrade to GPT-6 Sol for agentic coding, computer use, and professional work. Its model announcement lists API ID gpt-6.1-sol and ChatGPT Work/Codex access for Plus, Pro, Business, Enterprise, and Edu; ordinary Chat conversations are not yet supported. The list does not guarantee an identical model picker immediately in every account. Consult model and pricing pages for prices, context, and parameters rather than treating relative conference pricing as a billing quote.

For an established Sol workflow, save its model ID, sample inputs, tool permissions, acceptance criteria, and actual costs. Compare the same tasks in an isolated environment for correctness, latency, and total cost. Do not send a write-enabled agent into production by changing only a model name. The GPT-6 Sol and Luna guide separates earlier access and billing; older prices are not copied here as 6.1 prices.

Codex cloud, CLI, review, and security are separate features

The recap announces access to cloud tasks from computers, phones, and other devices, with reusable development environments for consistent settings and permissions. Listed plans are Plus, Pro, Business, Healthcare, Education, and Enterprise. The refreshed CLI adds voice operation, /agents, and session and worktree improvements for all plans. Cloud and CLI are not one installation step; inspect desktop, CLI, and cloud access separately.

Code review adds project diffs and reviews in the ChatGPT desktop app plus automated cloud review, labeled for all plans. Codex Security Cloud is a separate security product scanning GitHub repositories on demand or on a schedule. Listed access is Pro, Business, Enterprise, and Edu, including Daybreak Blue without a separate Daybreak application. Verify account entry points, repository authorization, and review results; the announcement itself is not a completed security check.

Limited preview is not universal API availability

Decisions API uses text or image context to select predefined answers for classification, routing, or an agent's next step. The recap explicitly describes a limited preview that day, with wider access planned in subsequent days. It does not establish permission for every API project. Prepare answer sets, error samples, and human-review criteria, then make real requests only when console and documentation establish eligibility.

Agents API announces computer use, multi-agent support, tool search, tool calls, and context compaction. The recap describes API availability alongside Pro 500 and Enterprise access in Codex and ChatGPT Work; these are separate product surfaces. For software-operating agents, restrict the test environment and tools, record observable results, then decide whether to allow real writes. The AI API configuration guide covers API Key, Base URL, and Model basics.

Dots, extensions, and AWS agents need separate eligibility checks

Dots is OpenAI's product for ongoing delegated work. The recap lists Pro and Business Premium in eligible markets. Enterprise, Edu, and Healthcare beta access is disabled by default and requires workspace-administrator enablement. When the entry point is missing, check plan, market, and settings before blaming connectivity.

Plugin extensions provide native experiences in ChatGPT's sidebar and interactive panels. OpenAI also announced support for the proposed MCP Events specification, allowing connected-app events to trigger automations. These differ from Dots in eligibility and task model. OpenAI and Amazon also announced Bedrock Managed Agents powered by OpenAI on AWS resources. Check AWS preview regions, permissions, and prices; OpenAI API availability does not automatically authorize a Bedrock account.

Turn an announcement into a verifiable decision

Choose one update relevant to existing work rather than migrating all models and agents at once. Record account, plan, workspace policies, region, target entry point, and official access scope. Keep limited previews and future features on a watch list. Once accessible, verify permissions, responses, errors, and actual costs using examples without production side effects. Restrict repository, browser, or external-app writes and retain human acceptance.

For each trial, save the original configuration, representative tasks, quality and cost baseline, and stopping conditions. For missing models, denied access, or quota notices, check eligibility and parameters first. Networking comes later, with DNS, TLS, timeout, or proxy 407 evidence. See the connection failure checklist. This is an announcement-scope and verification guide, not a claim of hands-on testing.

Sources

Frequently Asked Questions

Why do Chinese reports say September 30?

DevDay took place September 29, 2026 in US time, with broadcasts and announcements reaching September 30 in Beijing time. The difference is time zones; verify product facts against official announcements.

Is GPT-6.1 Sol available to everyone?

Not at every entry point. OpenAI announced API ID gpt-6.1-sol and ChatGPT Work/Codex for Plus, Pro, Business, Enterprise, and Edu. Ordinary Chat conversations were not yet supported; check workspace access.

Was Decisions API fully open?

No. The recap calls it a limited preview with expansion planned afterward. Check actual project eligibility and official documentation.

Does Agents API access grant Dots or Bedrock Managed Agents?

No. Verify the API project, ChatGPT plan and workspace settings, and AWS region and permissions separately.

What should a team try first?

Choose one relevant update, verify eligibility and documentation, then compare features, quality, and cost without production side effects. Track limited previews and retain confirmation for agents that write.