Apart from serving the ChatGPT experience to hundreds of millions of users everyday, OpenAI is also engaged in providing its platform for developers creating AI applications. OpenAI is popular among developers not just for its cutting-edge models, but also because of its strong tooling and support for developers.
Today, OpenAI announced two significant updates for developers. The first is about Codex, OpenAI’s software engineering agent. OpenAI is now making Codex available to ChatGPT Plus users. For a limited time, ChatGPT Plus users will be able to enjoy generous usage limits, but OpenAI will rate-limit them during high-demand periods.
Codex can now connect to the internet to install dependencies, upgrade packages, run tests that need external resources, and more. OpenAI specified that internet access is off by default, but users can enable it for specific environments. Users can also control the specific domains which Codex can access and more. This Codex Internet access capability is available for ChatGPT Plus, Pro, and Teams users, and it is coming soon to Enterprise users.
With today’s update, Codex users can now update existing pull requests when following up on a task. Finally, users can now dictate tasks to Codex.
Apart from the above and bug fixes, OpenAI has made the following improvements to Codex:
- Added support for binary files: When applying patches, all file operations are supported. When using PRs, only deleting or renaming binary files is supported for now.
- Improved error messages for setup scripts.
- Increased the limit on task diffs from 1 MB to 5 MB.
- Increased the limit for setup script duration from 5 to 10 minutes.
- Polished GitHub connection flow.
- Re-enabled Live Activities on iOS after resolving an issue with missed notifications.
- Removed the mandatory two-factor authentication requirement for users using SSO or social logins.
The second major update from OpenAI today is about voice agents. OpenAI’s Agents SDK is now available in TypeScript and it comes with support for handoffs, guardrails, tracing, MCP, and other core agent primitives. This SDK also includes new support for human-in-the-loop approvals, allowing developers to pause tool execution, serialize and store the agent state, approve or reject specific calls, and resume the agent run.
OpenAI also released an updated speech-to-speech model with improvements in instruction-following reliability, tool-calling consistency, and interruption behavior. Also, developers can now customize how fast the voice speaks during each session. Developers can now access the updated model via gpt-4o-realtime-preview-2025-06-03 in the Realtime API and gpt-4o-audio-preview-2025-06-03 in the Chat Completions API.
Finally, the Traces dashboard now supports Realtime API sessions, allowing developers to easily visualize voice agent runs, including audio input/output, tool invocations, and interruptions.
0 Comments - Add comment