See How Visible Your Brand is in AI Search Get Free Report

OpenAI DevDay 2025 Reveals GPT-5 Pro, Apps, and AI Agents — Is ChatGPT Becoming a Platform?

  • October 7, 2025
    Updated
openai-devday-2025-reveals-gpt-5-pro-apps-and-ai-agents-is-chatgpt-becoming-a-platform

OpenAI outlined a platform play at DevDay 2025: apps inside ChatGPT, a production AgentKit, and API updates, including GPT-5 Pro and Sora 2 access.

OpenAI also shared scale metrics that set the context for this push, with 4M developers, 800M+ weekly ChatGPT users, and 6B tokens per minute on the API.

📌 Key Takeaways

  • Apps in ChatGPT ship with a preview Apps SDK to build interactive, in-chat experiences.
  • AgentKit lands for building production-grade agents, alongside Codex GA.
  • GPT-5 Pro becomes the high-accuracy API model for critical workloads.
  • Sora 2 video generation is available via API in preview.
  • OpenAI highlights a growing MCP ecosystem for connectors and tools.


What OpenAI Announced

OpenAI is moving ChatGPT from answers to actions, starting with apps that run directly inside the chat window and an Apps SDK so developers can ship their own.

The site lists launch tiles for Apps in ChatGPT, AgentKit, Sora 2, Codex, GPT-5 Pro, gpt-realtime-mini, and gpt-image-1-mini.

The framing is about reach and reliability. With 4M developers building on the platform and 800M+ weekly users, OpenAI is tying agentic tools and MCP connectors to a massive surface where UI and conversation live in one thread.

“It’s the best time in history to be a builder.” — OpenAI


Apps In ChatGPT And The SDK

Inside ChatGPT, users can call apps by name or accept smart suggestions, then interact through embedded cards like maps, lists, and players. The Apps SDK arrives in preview, giving developers a path to ship interactive, context-aware tools that run in chat.

Under the hood, the SDK leans on Model Context Protocol (MCP), which standardises how models access tools and data. That lets builders reuse connectors, manage permissions, and keep logic portable across environments.


Agents, Codex GA, And Connectors

AgentKit targets production agents with tooling for workflows, evaluation, and admin controls. In parallel, Codex moves to general availability, with docs and controls highlighted on the DevDay hub.

The message is consistent: agents must be observable, governable, and shippable.

OpenAI points developers to connectors and MCP servers for files, storage, and third-party systems. This consolidates data access behind standard endpoints and reduces ad-hoc function plumbing, a common adoption barrier for agentic apps.

“Build with our toolkit for production-grade agents.” — OpenAI

Interactive apps in chat, AgentKit for shipping agents, Codex GA, and API access to GPT-5 Pro, Sora 2, realtime-mini, and image-1-mini.


API Models, Pricing, And Scale

Event materials call out GPT-5 Pro availability in the API for high-accuracy tasks, plus gpt-realtime-mini and gpt-image-1-mini for cost-efficient voice and image work. Sora 2 and Sora 2 Pro are listed in preview with aspect-ratio and sound controls.

The same materials detail token pricing for GPT-5 Pro, real-time and audio minis, and image generation, alongside per-second Sora 2 video pricing. Capacity metrics also step up sharply from 2023 to 2025, with the API now processing billions of tokens each minute.


Why It Matters

Putting apps in chat, standardising connectivity with MCP, and packaging AgentKit gives teams a straighter path from prototype to production. The platform unifies UI, actions, and context, which shortens user journeys and simplifies governance.

For developers, the incentive is distribution and speed. For enterprises, the draw is a managed surface with admin and safety controls where agents act predictably. The bets on pricing tiers and smaller models signal a push to make everyday use economically viable.


Conclusion

DevDay 2025 frames ChatGPT as a platform, not just a bot. Apps, agents, and APIs converge in one place, backed by scale and a clearer path to ship. If reliability and privacy hold, this becomes a daily workspace, not a demo reel.

The next proof points are app submissions, monetisation, and enterprise rollouts. If partner apps stay useful and connectors remain stable, the flywheel turns, and chat becomes the front door for real work.


📈 Latest AI News

7th October 2025

For the recent AI News, visit our site.


If you liked this article, be sure to follow us on X/Twitter and also LinkedIn for more exclusive content.

Was this article helpful?
YesNo
Generic placeholder image
Articles written 859

Khurram Hanif

Reporter, AI News

Khurram Hanif, AI Reporter at AllAboutAI.com, covers model launches, safety research, regulation, and the real-world impact of AI with fast, accurate, and sourced reporting.

He’s known for turning dense papers and public filings into plain-English explainers, quick on-the-day updates, and practical takeaways. His work includes live coverage of major announcements and concise weekly briefings that track what actually matters.

Outside of work, Khurram squads up in Call of Duty and spends downtime tinkering with PCs, testing apps, and hunting for thoughtful tech gear.

Personal Quote

“Chase the facts, cut the noise, explain what counts.”

Highlights

  • Covers model releases, safety notes, and policy moves
  • Turns research papers into clear, actionable explainers
  • Publishes a weekly AI briefing for busy readers

Related Articles

Leave a Reply