OpenAI AI Updates: September 30, 2026

1. GPT-6.1 Sol Brings Near-Astra Performance at One-Fifth the Price

OpenAI. OpenAI released GPT-6.1 Sol at DevDay, one week after GPT-6 Sol, positioning it as close to GPT-6 Astra on agentic coding, computer use, and professional work. API pricing is $2 per million input tokens, $0.10 cached input, and $10 output, compared with Astra’s $10, $1, and $50. The model (gpt-6.1-sol) is available in the API, in ChatGPT Plus, Pro, Business, Enterprise, and Edu, and in Codex and ChatGPT Work, and Codex CLI 0.159.1 makes it the default model in its bundled model catalog. Source

2. Dots Are Always-On Agents With Their Own Cloud Computer

OpenAI. Dots are proactive, persistent agents in ChatGPT that run on GPT-6 Astra, each with its own cloud computer and browser, and keep working toward assigned goals around the clock while learning from feedback. They reach more than 4,000 apps through plugins and can be used from the desktop and mobile apps, the web, Slack, and Teams. Dots are rolling out to Pro and Business Premium users in eligible markets, with Enterprise, Edu, and Healthcare in an admin-enabled beta; the first dot is included in the plan and chatting with it does not count toward usage limits. Source

3. Ultrafast Tier and a New Pro 500 Plan

OpenAI. A new Ultrafast processing tier runs GPT-6 Astra at up to 300 tokens per second, which OpenAI describes as up to 8x faster in Codex and 6x faster in the API, at six times the standard price ($60 input and $300 output per million tokens for Astra). Sol Ultrafast is listed as coming soon. The new Pro 500 consumer plan includes Ultrafast and 25x the Plus usage allowance; for new $200 Pro subscribers the usage multiplier drops from 20x to 10x, while existing subscribers keep 20x. Source

4. Agents API Adds Computer Use, and a Luna-Based Decisions API Enters Preview

OpenAI. The Agents API offers OpenAI-hosted agents with multi-agent controls, tool search, context compaction, memory, and now computer use. The new Decisions API, in limited preview with broader release promised in the coming days, restricts GPT-6 Luna to a developer-defined set of answers for classification, request routing, and agent action selection, at Luna’s $0.10 input and $0.50 output per million tokens. Source

OpenAI. Sign in with ChatGPT lets third-party apps authenticate Plus and Pro users and bill eligible model usage against their existing ChatGPT plan, with 16 launch partners including Cognition’s Devin, Notion, and Vercel. Plugins gain extensions such as sidebar panels and file viewers, and the plugin spec adds MCP Events. The OpenAI Marketplace lets enterprises apply OpenAI spend to partner software, with launch partners including Adobe, Figma, Harvey, CrowdStrike, and Baseten. Source

6. ChatGPT Space and Pages Target Team Collaboration With Agents

OpenAI. ChatGPT Space is a shared team home for files, documents, and scheduled tasks where people work alongside their dots, launching on Pro, Business, and Enterprise desktop and web. Pages is a new document type for human and agent co-editing, collaborative Slides with PowerPoint and Google Slides export is due in the coming weeks, and a beta Meetings plugin for the macOS app deletes audio once notes are generated. Source

7. Codex Gets Cloud Tasks on More Plans, Automated Code Review, and Security Cloud

OpenAI. Codex cloud is now included in Plus, Pro, Business, Education, Healthcare, and Enterprise plans, and automatic Code Review covers GitHub pull requests and GitLab merge requests from the ChatGPT desktop app on all plans. Codex Security Cloud, for Pro, Business, Enterprise, and Edu, scans GitHub repositories on demand or on a schedule, de-duplicates findings, and prepares patches as PRs. Codex CLI 0.159.0 adds an opt-in instant_interrupt setting so new input can steer Codex mid-response, plus better Mermaid rendering and paginated thread history for app-server clients. Source

8. OpenAI Publishes Early Guidelines for Safety Cases in Frontier Training

OpenAI. Published late on September 28, OpenAI’s safety post sets out early guidelines for building safety cases around frontier model training runs. The guidelines cover technical safeguards, operational practices, and how to investigate misalignment incidents, following the agent incidents and training pause the company disclosed in recent days. Source