We’re excited to share some important updates to our image and PDF APIs that delivers faster speeds, better multilingual support, and a more dependable experience overall.
Faster processing times
We’ve increased processing speeds by 25% resulting in quicker turnaround, even during periods of peak demand.
This reduction means less waiting, more responsiveness, and a smoother experience end-to-end.
Our most reliable PDF API yet
In addition to faster speeds, we’ve made several major architectural and monitoring upgrades to the PDF API. As of July 22nd, the PDF service is performing at its most consistent level yet.
Thanks to enhanced observability, more aggressive external monitoring, and new internal instrumentation, we’ve delivered:
- Steady reliability, even under fluctuating demand
- Fewer performance spikes and smoother throughput across all workloads
These upgrades reflect a focused effort to strengthen the reliability of our OCR service. We know how important dependable processing is to your workflows, and we’ve built in the tooling to ensure we stay ahead of any issues. We’ll continue to actively monitor and invest in platform stability to ensure that your experience remains seamless at any scale.
To view the current status of our APIs, you can always visit https://monitor.mathpix.com
Improved model performance
In addition to backend improvements, we’ve also upgraded our core model. The latest model release brings significantly better inference quality for Gujarati, Tamil, and Bangla along with broader accuracy gains across all of our supported document types, helping reduce misrecognitions and edge-case failures.