What's New in ChatGPT for Developers

A short refresher on the latest ChatGPT, Codex and OpenAI API updates for developers

What this course covers

The course runs across 7 topics, each one a short adaptive session rather than a recorded lecture. The tutor explains the idea, works an example, checks understanding, and adjusts the next step to the answer given.

GPT-5.5 removed from Codex — What's different: GPT-5.5 stopped working in Codex for people signed in with ChatGPT on October 14, 2026; API-key sessions are not affected. Why it matters for your work: Saved settings, custom agents or scheduled tasks that still pick GPT-5.5 need a new model or they will not run as expected. Try it (about 2 minutes): Open the model picker and choose GPT-6.1 Sol instead, then check any saved settings or scheduled tasks that named GPT-5.5. Needs: the Codex app. No access? Here's what you'd see: Users signed in with ChatGPT will find GPT-5.5 no longer works in Codex, but they can select GPT-6.1 Sol instead. Renamed or removed: GPT-5.5 removed from Codex (ChatGPT sign-in); use GPT-6.1 Sol
GPT-6.1 Sol is the new main Codex model — What's different: OpenAI added GPT-6.1 Sol to Codex, a model that comes close to its top model on hard work at a lower cost, and made it the default model in the Codex CLI. Why it matters for your work: It is now the recommended choice for long coding and document jobs, so your tasks may run better without you changing anything. Try it (about 2 minutes): Open the model picker in the Codex app (or type /model in the CLI) and check that GPT-6.1 Sol is selected, then run a small task with it. Needs: the Codex app. No access? Here's what you'd see: GPT-6.1 Sol appears as the default model in the Codex CLI and is available in the Codex app model picker.
Older API models were shut down — What's different: On October 23, OpenAI turned off API access to a set of older models, including older GPT-3.5, GPT-4, GPT-4 Turbo and GPT-4o snapshots, o1, o1-pro, o3-mini, o4-mini and gpt-image-1. Why it matters for your work: Any app or script that still calls one of these models now gets errors until it is moved to a newer model. Try it (about 2 minutes): Search your code and settings for the old model names, then open the deprecations page and swap each one for the replacement it lists. Needs: the OpenAI API. No access? Here's what you'd see: Apps and scripts calling shut down legacy OpenAI API models, such as older GPT-3.5 and GPT-4 snapshots, now fail with errors. Renamed or removed: Legacy API model snapshots (GPT-3.5 Turbo, GPT-4, GPT-4 Turbo, GPT-4o 2024-05-13, o1, o1-pro, o3-mini, o4-mini, gpt-image-1) shut down
Assistants API shut down — What's different: The Assistants API was shut down on August 26, 2026. Why it matters for your work: Any app still calling the Assistants API will fail and must move to the Responses API. Try it (about 2 minutes): Search your code for Assistants API calls (threads, runs, assistants) and use the Responses API or Agents API instead. Needs: the Assistants API. No access? Here's what you'd see: Calls to the shut-down Assistants API will fail, and developers must use the Responses API or Agents API instead. Renamed or removed: Assistants API removed; use the Responses API (or Agents API) instead
New GPT-6 model family in the API — What's different: The API added a GPT-6 family, starting with a flagship model on September 3 and followed by a cheaper all-round model and a small fast model on September 22. Why it matters for your work: Apps built on older models may get better results or lower costs by switching, but the new models drop some settings like custom temperature. Try it (about 2 minutes): In the Platform playground, run one of your real prompts on a GPT-6 model next to your current model and compare the answers. Needs: the OpenAI API.
Also new — - Import your setup from Claude Code, Cowork or Cursor: The desktop app and CLI can now import setups from Claude Code, Cowork, and Cursor. - GPT-5.4 and GPT-5.4 mini removed from Codex: GPT-5.4 and GPT-5.4 mini have been removed from Codex for ChatGPT-linked sessions. - Steer the CLI while it is still answering: The Codex CLI now allows users to send follow-up prompts to steer active responses. - Retirement of the 'untrusted' approval setting: The untrusted approval setting for sandbox permissions has been retired. - Search and change CLI settings with /config: The Codex CLI now includes a searchable /config panel to adjust settings directly. - GPT-6 Luna for quick, high-volume tasks: GPT-6 Luna is now available in the desktop app for quick, high-volume tasks. - Agents API in public beta: OpenAI released the Agents API in public beta with managed sandboxes and streaming. - Codex cloud now works with GitLab (beta): Codex cloud now supports GitLab integration for tasks and merge request reviews. - The --full-auto flag is gone from codex exec: The --full-auto flag was removed from codex exec in favor of new sandbox flags. - Codex Cloud environments now require publishing and support desktop reopening: Codex Cloud environments now require publishing before they can be reopened in the desktop app. - and 12 more: GitHub pull request review transitioned to Codex Cloud (Legacy); New memory system is on by default in the CLI; Worktrees are on by default in the CLI; AGENTS.md only loads in projects you trust; Select and search the CLI transcript by keyboard; Built-in Google Cloud Vertex AI provider; Limit which service tiers each project can use; codex mcp-server command removed; Clearer explanations when a request is blocked; Change reasoning effort mid-conversation; Decisions API is now generally available; GPT-6 Rosalind for approved life sciences labs
Try one on your own work — Pick the one main change from this course that is most useful to your job. Try it in your own ChatGPT on a real task, then paste or describe what you got. The coach gives short, specific feedback. No access to the app or plan it needs? Describe how you would use it on a real task this week. That counts too. There is no score and no pass or fail: finishing this completes the course.

Skills you build

Results are measured against named skills in the OneRange taxonomy of more than 10,000 skills, so a manager sees proficiency per skill rather than a completion tick.

Who it is for

The material is pitched at intermediate level, and takes roughly 20 minutes at a typical pace. Because every session adapts, someone who already knows a topic moves through it quickly instead of sitting through an explanation they do not need.

It sits in the Tool: ChatGPT track of the Vero AI Catalog, and can be assigned to one person, a team, or the whole company.

How it is delivered and assessed

Delivery is conversational and interactive. There is no video to sit through and no slide deck to click past.

Understanding is checked with a lab assessment, with a pass mark of 70%.

Administrators can copy this course into their own library and adapt it — edit the outline, change the duration, swap the assessment format, or ground it in internal documentation so answers cite the company's own source material.

Also in this track

All ChatGPT training courses →
ChatGPT: Core — Custom GPTs, Data Analysis & Reasoning — A hands-on working level for experienced ChatGPT users in any role. Build Custom GPTs and Projects, analyze uploaded data, compare reasoning approaches, bring your apps into Work through plugins and @mentions, and run Work tasks with a plan and approvals. No technical background needed; for code and integrations take ChatGPT for Developers.
ChatGPT for Developers: Core — Codex in ChatGPT & MCP — Codex for working developers, where it now lives: inside the ChatGPT desktop app as Codex mode, plus the Codex CLI and IDE extension. Learners set up a project, give Codex the right context with AGENTS.md, choose how much it may do without asking, run planned coding tasks on a repository, review diffs and pull requests with discipline, connect tools with MCP, and work in large codebases. Ends with a ticket taken from plan to reviewed change.
What's New in ChatGPT for Everyone — A short refresher on the latest ChatGPT for Everyone updates
ChatGPT: Advanced — Power Workflows & Automation — The step between everyday fluency and leading work with ChatGPT, without writing code. Learners design multi-step workflows, automate recurring work with Work, plugins, Scheduled tasks, Custom GPTs and Sites, keep long-running tasks on track, and set personal systems that compound over weeks. They finish as the person on the team others come to for ChatGPT technique.
ChatGPT: Essentials — Data & Analysis Playbook — A focused playbook for doing data and analysis work in ChatGPT. Learners run real tasks end to end - data cleanup, analysis, charts, and interpretation - and leave with repeatable workflows they can use the same day. About 45 minutes including the final.

Frequently asked questions

How long does this course take?

It runs roughly 20 minutes at a typical pace. Because every session adapts, someone who already knows a topic moves through it quickly instead of sitting through an explanation they do not need.

How is it assigned to a team?

An administrator assigns it from the Vero dashboard to one person, a team, a department or the whole company, and sees progress and results per person and per skill.

Can we customise it with our own documents and terminology?

Yes. Administrators can copy this course into their own library and adapt it — edit the outline, change the duration, swap the assessment format, or ground it in internal documentation so answers cite the company's own source material.

Course code what-s-new-in-chatgpt-for-developers-e3d2f2fb.