Introducing TEO-3.5, Our Most Reliable Model Yet
Release - April 2026
Our AI model is getting a huge upgrade. It will continue to monitor the residents just like before, but it is now significantly more reliable.
Key improvements
These fundamental improvements in the model's understanding of the room lead to more timely alarms, improved health statistics, and increased safety for residents and patients.
38% fewer errors
when detecting whether someone is in the room. For you, this means fewer false alarms.
Resident
Staff31% fewer misclassifications
of residents, staff, and visitors. For you, this means a better overview.
43% fewer missed detections
when a resident is in bed. For you, this means fewer false alarms while the resident is sleeping.
On the technical side
Three major reasons TEO-3.5 is special
New architecture
We use a state-of-the-art architecture called a transformer. This is the same type of technology used in systems like GPT and Google's Gemini. This improved design means that the model learns and generalizes better.
Much more training data
TEO-3.5 is trained on more than 1.7 billion images. This is more than 5,600x the data used for our previous model.
No handcrafted heuristics
Instead of relying on manually designed rules to filter detections, the model now learns to reason about objects directly from data. This means more stable and reliable results.
Same experience
While the technology behind TEO-3.5 is significantly more advanced, your experience remains the same. Just more reliable. The system will continue to monitor residents and patients as before, but now with fewer false alarms and more accurate detections.