Models & providers

Choose your provider

OpenAI, Anthropic, Google Gemini, AWS Bedrock, SpaceXAI, Meta, OpenRouter, Microsoft Azure, Zai, Alibaba, Nvidia, Deepinfra, Mistral, DeepSeek, Moonshots / Kimi, and TogetherAI.

The signed-in workspace provides editable connection presets and a model dropdown loaded from your provider account. Availability depends on your provider key. Azure deployments and models without a catalog can use an exact model ID.

Your endpoint. Your model.

Custom connections support HTTPS gateways, authentication headers, JSON request templates, and response-text mappings. OpenAI-compatible servers can host open-weight models. Private-network endpoints and custom agents can use the downloadable local benchmark core.

Choose a provider or configure a custom endpoint →

Publication status

Ten current-release runs are internally reviewed. Their measured scores and ranks are public under anonymous model classes; direct identities remain withheld until external validation.