Working with OpenAI's o3, o4-mini, and GPT 4.1 in Xcode using Alex Copilot

Today, we’re releasing OpenAI o3, o4-mini, GPT 4.1, and Gemini 2.5 Pro (without API key) in Alex Sidebar! You can use these models with our agent to generate Swift, iOS, and MacOS apps with Xcode.

OpenAI o3

We’ve been super impressed with this model. Definitely state of the art.

Note: This model uses o1 credits, which you need to purchase separately in the billing settings. Or you can use your own OpenAI API key if you have access to o3 (check https://platform.openai.com/settings/organization/limits).

OpenAI o4-mini - Fast, lightweight reasoning model.

OpenAI GPT 4.1 — New, fast, and pretty good alternative to Claude 3.5 Sonnet.

Gemini 2.5 Pro — You can now also use Gemini 2.5 Pro without your own API Key! It counts as a normal chat message.

As always, using your own API Keys lets you bypass the chat message limits for all models. Have fun!

Get started here.

A note on o3

From my personal experience, o3 has been a huge step forward compared to other models. Its ability to use tools has skyrocketed since o1/o3-mini. It's now comparable to (or even beating) Claude 3.5 Sonnet at tool use.

A big problem o1 had was it would hallucinate the existence of files. It really struggled with determining if a file was included in its context, which resulted in extreme hallucinations.

Now, the o3 agent doesn't seem to run into this issue, and can reason over what tools to use. This is huge.

Of course, it's still an extremely expensive model, and doesn't stream its thinking tokens, but still a huge leap for code quality and debugging.

If you're an iOS dev, or use Xcode for your daily work, check it out!