Changelog

v2.8.2: Opus 5.5 by Name, a Refreshed Codex Line-Up, and a Model List That Updates Itself

Model line-ups moved twice in a fortnight. Anthropic shipped Opus 5.5, OpenAI reshuffled its Codex range, and this release brings the model picker back in step with both.

Opus 5.5, by name

If you pick Opus for a Claude employee, you have been running Opus 5.5 since Anthropic released it. The picker just kept calling it Opus 5.

It now says what you are actually getting. The rest of the Claude line-up was already current: Fable 5.1, Sonnet 5 and Haiku 4.5.

A refreshed Codex line-up

For Codex employees the picker reads your account's own catalogue, so it normally shows exactly what your account can use. There is a built-in list behind it for the moments when the catalogue cannot be reached, and that list had drifted.

It is current again. gpt-6-astra is the new default. Two retired models are gone, gpt-5.2 is back where it belongs, and the reasoning levels match what Codex actually offers today: none and minimal have gone away, while max and ultra are now available.

The part that matters beyond this release

Until now the default model was pinned by name in our own code. Every time a provider changed its default, we had to ship a release to catch up, and in between you would see a name that was quietly out of date.

Model selection now follows the catalogue's own default instead. When your provider promotes a new default model, your employees pick it up on their own. No release, no waiting, no stale name in a dropdown.

That is the difference between keeping a list and reading one, and it is why this is the last time you will read a changelog entry about a model name being out of step.

Fresh runtime under the hood

The agent image was rebuilt with the current tooling: Claude Code 2.1.280, Codex CLI 0.156.1 and OpenCode 1.18.32.

Nothing to do on your side. Open the picker and you will see the right names, with the right reasoning levels, and the right default already selected.

Want to test the most advanced AI employees? Try it here: https://geta.team

v3.1.1: Give Every Conversation Its Own Background, German and Hindi Join the Interface, and Email Cards Stop Filling the Screen

Two languages, a lot of colour, and an email card that no longer swallows your screen. Every conversation can have its own background If you have ever set a wallpaper in WhatsApp, you already know what this is. Open Configure, then Chat background, and pick the backdrop for your conversation

v3.1.0: The Geta.Team Interface Now Speaks Eight Languages, Incoming Email Stops Eating Your Budget, and an Export Finally Carries the Whole Employee

Three big things in this one. The app itself now works in your language, incoming email stops quietly eating your budget, and an exported employee finally arrives somewhere else still being itself. One thing worth being clear about before we start: your employees already speak around a hundred languages, and

v3.0.0: Video Call Your AI Employees, Change Their Look Live, and Full OAuth Support for MCP Servers

Version 3.0.0 is the first major version number we've shipped in a while, and it earns it. Your AI employees can now pick up a video call, look you in the eye while they talk, and change outfits when you ask. Underneath the showpiece there'

How to Hire an AI Employee: A Step-by-Step Guide for Small Business Owners

Most small businesses that give up on an AI employee don't give up because of the AI. They give up because they skipped steps they would never skip with a person: no job description, no introduction to the team, no idea who the new hire is allowed to