Mistral vs ElevenLabs: Is the Voice AI Revolution French? π«π·
Hey everyone,
Last week, Mistral released a new Text-to-Speech model (Voxtral) capable of running 100% locally, without sending any data to the cloud. I wanted to test it head-to-head against ElevenLabs β and honestly, the results are really good!
It complements their Speech-to-Text model released back in February. I put both to the test, and you can see (and hear) the comparison in my latest video:
A change in Gladys Plus: moving to Mistral Small 3.2 on Scalewayβ
While exploring Mistral's models, I'm also announcing a change to the AI powering Gladys Plus.
I'm replacing OpenAI GPT-5 mini with Mistral Small 3.2, hosted in France on Scaleway π«π·
The model is faster, cheaper, and your data stays with a sovereign French host. It's win-win-win! In concrete terms, this affects:
- camera image analysis
- proactive scenes
- conversations with the AI
And since the model is cheaper to run, I'm taking the opportunity to increase the monthly quotas included in the subscription:
- 3,000 text requests / month
- 500 camera image analyses / month
My goal is to unlock new use cases for you. If AI prices drop by another factor of 10 in the coming months, I'll raise the limits again π
If you haven't tried Gladys Plus yet, there's a free trial:
π Try Gladys Plus

