Reflection AI's Beam now available on AI Gateway for free

Reflection AI’s Beam model, previously known as stealth/cat-jam, has been introduced to AI Gateway, making it available for immediate use. The model is currently free to access during a limited-time launch period when served by Reflection AI, though usage may incur costs from the provider after the promotional period ends. Beam is designed for coding and agentic tasks, featuring tool calling, structured outputs, and a 1-million-token context window. Its architecture prioritizes speed by activating only 23 billion of its 501 billion parameters per token, reducing unnecessary computation and accelerating response times compared to larger open models.

To integrate Beam, users must employ the standard model name, reflection/beam with the provider reflection, or the free variant reflection/beam-free during the promotional window. Once the free period concludes, the model slug will transition to paid pricing without service interruption, though the -free version will no longer be servable. The transition ensures continuity for existing workflows while shifting to standard billing practices. Developers are advised to update their configurations accordingly to avoid disruptions after the promotional offer expires.

Beam can be deployed in coding environments such as Claude Code, Codex, and Cursor by installing the latest Vercel CLI and running the setup command. The process involves selecting either reflection/beam or reflection/beam-free as the agent’s model, depending on the current availability period. Vercel’s documentation provides step-by-step guidance for integrating the model into these tools, ensuring compatibility with existing development workflows. This integration aims to streamline the adoption of Beam for real-time coding and agentic applications.

AI Gateway serves as a unified interface for accessing Beam and other models, offering features like usage tracking, cost management, and performance optimizations. The platform includes Zero Data Retention support, custom reporting, and routing rules to enhance reliability and uptime. AI Gateway mirrors provider pricing without additional markup and does not impose platform fees on inference, including for Bring Your Own Key requests. This structure ensures transparent and cost-effective access to Beam and other supported models.