LM Arena

LM Arena provides a web-based environment for running, comparing, and deploying large language models. It focuses on making model evaluation, hosted inference, and simple deployment workflows accessible from a browser and via an API. The platform supports uploading or connecting models, running evaluation workloads, and exposing inference endpoints for applications.

• Model comparison run side-by-side evaluations and benchmarks across models.
• Hosted inference deploy models to managed endpoints for production usage.
• API access programmatically invoke models and integrate into applications.
• Custom model uploads bring your own model artifacts for testing and deployment.
• Usage monitoring track usage metrics and performance of deployed endpoints.

To get started, sign up on the web app, upload or connect a model, run a comparison job, and create an inference endpoint; use the provided API keys to integrate inference into your applications.