Gladys Plus switches to ChatGPT 4.1-mini: faster, smarter!

[pierre-gilles]

Hello everyone

OpenAI released a new model last month: ChatGPT 4.1-mini, announced as smarter and faster.

On my side, I have a suite of automated tests to evaluate model quality in a smart home context. And good news: in these tests, 4.1-mini gives better results than the 4o-mini model we had been using until now.

I’ve therefore updated Gladys Plus to use this new model right away

If you notice strange responses or cases that don’t work, feel free to send them to me: I’ll add them to my test suite to continue improving quality.

To join Gladys Plus: Gladys Plus | Gladys Assistant
More info on ChatGPT 4.1 here: https://openai.com/index/gpt-4-1/

@PhilippeMA For the example we were discussing this afternoon, switching to ChatGPT 4.1-mini does fix the problem on my side

[PhilippeMA]

Hello,

Great, very responsive! I tested it too and it’s fine.

Thanks

[pierre-gilles]

I also spotted a small bug in sending the chat history to ChatGPT (used as context to improve the relevance of responses).

This bug is now fixed in this PR ChatGPT: When asking the AI, should send better context by Pierre-Gilles · Pull Request #2316 · GladysAssistant/Gladys · GitHub

The fix will be available in the next version of Gladys

[pierre-gilles]

GPT-5 is out!

According to the benchmarks published by OpenAI, performance should be much better than GPT-4.1 that we’re currently using in Gladys Plus

Fewer hallucinations, supposedly:

I’ll run some tests, and if it’s conclusive, we can move to this new model!