Lavita AI’s Medical Sphere is the go-to platform for real-time evaluation of foundation models on any medical or clinical task, at scale. We’re building a global network of medical professionals, alongside a growing community of users, to create the most trusted and comprehensive ecosystem for medical AI.

Lavita AI’s Medical Sphere

​
How It Works

On the Medical Sphere, users can evaluate and test various AI models on medical tasks. They can chat with two foundation models and compare their performance based on criteria such as correctness and helpfulness. Everything begins with asking a medical question. Users can engage in either a single-turn or multi-turn conversation, then compare the responses from both models and vote for one of the following options:
  1. Model A is better
  2. Model B is better
  3. It’s a tie (both models performed well)
  4. Neither (neither model provided a good response)
Medical Sphere

Medical Sphere

​
Comparison Modes

There are two modes of comparison on the Medical Sphere. By default, the models remain anonymous to ensure an objective comparison. However, users can choose to switch off the “Anonymous” option and manually select which models to compare—this is called a non-anonymous comparison. In anonymous mode, after submitting a vote, the model names will be revealed.
Model selection in the non-anonymous mode

Selecting models in the non-anonymous mode

​
Live Leaderboard

Medical Sphere features a live leaderboard where users can track the real-time performance of AI models. To ensure fairness, rankings are based only on anonymous votes - cases where the model’s identity is hidden from the evaluator ,  helping eliminate bias in the evaluation process. We also actively monitor and review voting logs to filter out non-medical queries and instances where model identities are revealed in the responses. The leaderboard is periodically refreshed based on this cleaned, high-quality data , so what you see reflects meaningful, trustworthy feedback from the community.
When reporting results on our leaderboard, we only consider votes from anonymous battles. Additionally, while users can ask any type of question, we filter out votes on non-medical or non-clinical conversations when aggregating results. Therefore, we encourage users to focus on medical questions, as votes on non-medical topics will not be counted.