# Medical Sphere

> Medical Sphere is a community-driven platform where healthcare professionals and AI researchers evaluate, compare, and benchmark large language models on real-world medical tasks. It acts as a trust layer for healthcare AI — surfacing which models perform best on clinical reasoning, medical imaging, diagnostic support, and patient communication.

Medical Sphere combines an AI evaluation arena, a collaborative clinical case repository, and a standardized benchmarking suite into one platform designed specifically for the healthcare domain.

All AI model rankings are derived from expert votes using an ELO rating system, not automated metrics alone. Users can compare models anonymously (blind battle), head-to-head with known identities, or query a single model directly.

Medical professionals can apply for verification, which grants access to restricted models and marks their contributions with a "Verified" badge for clinical credibility.

The platform integrates models from OpenAI (GPT series), Anthropic (Claude series), Google (Gemini series), Meta (Llama series), DeepSeek, Mistral, xAI (Grok), and others. Multimodal analysis — including medical images (X-rays, MRIs, CT scans) and video — is supported for applicable models.

A public X/Twitter AI agent/bot (@AskMedSphere) allows anyone to mention the account with a medical question or image and receive an AI-generated response synthesized from multiple models.

## Core Platform

- [Homepage](https://medicalsphere.ai/): Overview, top-ranked models, and quick entry points to the arena and cases
- [Medical AI Arena](https://medicalsphere.ai/arena): Side-by-side model comparisons with blind battles, known comparisons, and single-model mode
- [Healthcare Leaderboard](https://medicalsphere.ai/leaderboard): Community-ranked model leaderboard based on expert votes and ELO ratings
- [Clinical Cases Feed](https://medicalsphere.ai/cases): Community-submitted clinical cases with AI insights, image annotations, and expert discussion threads
- [Healthcare Benchmarks](https://medicalsphere.ai/benchmarks): Standardized benchmark results including NEJM Bench, MedCalcBench, and MedAgentBench

## Key Concepts

- **Anonymous Battle**: Users evaluate two anonymous AI models on a medical query and vote for the better response — eliminating brand bias
- **ELO Rating**: Model rankings update in real time based on head-to-head outcomes using the ELO chess rating formula
- **AI Insights**: For each clinical case, multiple AI models run automatically and produce structured diagnostic responses that are summarized into a consensus interpretation
- **Image Annotation**: Case contributors and reviewers can draw regions of interest (ROIs) directly on medical images and attach clinical notes
- **Medical Verification**: Healthcare professionals can submit credentials for review; verified status unlocks restricted models and expert-tier features
- **@AskMedSphere Agent**: A social integration that processes X/Twitter mentions, runs them through the multi-model pipeline, and replies with a synthesized AI response

## Optional

- [Add a Clinical Case](https://medicalsphere.ai/add-case): Submit a new clinical case with description, images, and selected AI models for analysis
- [Articles](https://medicalsphere.ai/articles): Research summaries and technical reports on AI model performance in healthcare
- [Register](https://medicalsphere.ai/auth): Create an account to participate in arena battles, submit cases, and vote on responses