Today, we are officially announcing that we are shutting down the BEval Studio evaluation platform. We made this decision to focus on our other products over the next few quarters, and because of changes in the company's vision. Evaluation is still offered as a service.
What's changing
BEval Studio launched in January 2026 as a hosted, self-serve platform. Teams sent their model outputs through our SDK or API, and the platform logged them, scored them and surfaced failures in a dashboard.
The self-serve platform is closing. Evaluation continues as a service that our team runs for you.
Why we made this decision
Over the next few quarters, we are focusing on the products we build and operate, bolder.fit and Scailor.
A self-serve platform requires continuous work on ingestion, dashboards, billing, onboarding and support. Alongside changes in the company's vision, we decided that this effort is better spent on our core products.
Evaluation as a service
The evaluation engine behind BEval Studio remains in use. With the service, we:
- Define what a good output looks like for your use case, together with your team.
- Set up deterministic verifiers, LLM judges across providers and rubrics written for your product.
- Run the evaluations on your model's outputs and walk you through the results.
We continue to use the same engine for our own products. The coaching agent in bolder.fit logs every model call to it.
If you use BEval Studio today
We will contact you directly with the shutdown date and instructions for exporting your data. If you would like evaluations to continue, we can move you to the service.
For any questions, contact us at support@bolder.services.
To discuss evaluation for your product, book a call.