Hey there,

OpenAI wrapped up their highly anticipated DevDay this week and have released a wave of new features for ChatGPT, their API, and more. These latest developments aim to empower users and developers alike with tools like Apps in ChatGPT and AgentKit. Today we’ll go through these announcements and highlight the changes with the biggest impact.

The Biggest Announcements from DevDay 2025 

This year’s DevDay was packed with all kinds of new models, tools, and features, but a few announcements stand out as the most noteworthy. Here’s what you need to know: 

Apps in ChatGPT: 

This new tool significantly improves the way you complete tasks and find answers within ChatGPT. Users can now “call” on different apps by starting the message with the name of an app. Each app has a unique use case, vastly expanding what ChatGPT can actually do. In other words, ChatGPT has become a new medium for interacting with third-party products using natural language. Right now there are only a handful of pre-built Apps, but developers have already started creating their own. In the example below, I ask ChatGPT to explain LLMs while calling on the Coursera app. The results were awesome!

ChatGPT will also connect you with relevant apps depending on what you are looking for. In this example, I ask ChatGPT about finding a fictional home in San Jose at a specific price range (one I can’t afford). 

After I click “Use Zillow”, it will regenerate its response and you’ll see the following.

One important note is that you have to manually download / connect to each app. Even if ChatGPT recommends you try one, it won’t start using it without manual confirmation from the user. 

 New Models: 

OpenAI has released several new models that are available via their API. I’ll give a brief summary of each model, but I recommend reading the API docs for pricing and detailed specs. 

• GPT-5 Pro: 

GPT-5 Pro is the most powerful reasoning model OpenAI has ever created. This model is meant to tackle complex, multi-step problems and can take up to several minutes to complete some requests. 

• gpt-realtime-mini:

This is a cheaper, light-weight version of the GPT Realtime model. It’s meant to be a fast, high-performance model with the capacity for understanding both audio and text inputs.  

• gpt-image-1-mini:

This model is effectively a cheaper version of the GPT Image 1 model. It can handle both image and text inputs while producing image-only outputs. 

• Sora 2: 

While Sora 2 has been out for a few months, it is now available within the Video API, allowing developers to build with OpenAI’s newest video generation model. This API allows developers to create, download, list, and delete Sora 2 videos using new endpoints. 

Codex Updates: 

Codex has received several big updates including three new features; Slack Integration, Codex SDK, and new admin tools. Here’s a breakdown:

• Slack Integration: Codex can now be linked directly to Slack channels. This means that you can interact with Codex as though it were another team member, assigning coding tasks and asking questions.

• Codex SDK: You can now add Codex directly to your engineering workflows, tools, and apps. 

• New Admin Tools: Developers now have greater control over workspaces and have access to new analytics dashboards. 

Overall, these new features transition Codex from a cool experimental tool to an integrated part of the development team. 

AgentKit: 

OpenAI has created a new tool for developing agents. AgentKit is meant to streamline the creation and deployment process for developers. The main improvements revolve around using “building blocks” like Agent Builder which is a visual canvas for creating agent workflows. This tool simplifies the process by including  “drag-and-drop nodes, connecting tools, and configuring custom guardrails”.

My Thoughts

OpenAI’s 2025 DevDay was definitely impressive. From the excellent new models to the significant ChatGPT upgrade, OpenAI didn’t disappoint. The changes to ChatGPT in particular piqued my interest. Giving people the ability to connect with apps not only helps them have a better experience, but it also keeps them within the ChatGPT ecosystem. I’m always impressed with how OpenAI is able to expand the reach of ChatGPT, first with plug-ins, then with custom GPTs, and now with these new apps. I can’t wait to see what inventive apps we see in the coming months. 

It’s also clear that OpenAI is prioritizing agent creation based on these latest changes. The AgentKit allows devs a smoother path for building agents. Whether you’re a developer excited to build with AgentKit or an everyday ChatGPT user ready to explore Apps, OpenAI has showcased changes that everyone should be excited about!