Beam's beta API is OpenAI-compatible
Model ID Beam-501B-A23B, 256K context and 128K output, available to early-access users.
Reflection's developer documentation lists one model, Beam-501B-A23B, with a June 30, 2026 knowledge cutoff, a 256K-token context window and up to 128K output tokens. The docs note the context window may change during the beta.
Besides its native API, Reflection offers an OpenAI-compatible endpoint at https://api.reflection.ai/openai/v1. It supports Chat Completions and model listing, so the official OpenAI SDKs work after changing the base URL. Responses, Embeddings, Files and Batch are not supported.
Beam supports reasoning, tool calling and structured outputs. Reasoning is always on, with an effort setting to trade speed for accuracy. No price has been published for the beta.
Takeaway
If you already use the OpenAI SDK, switching to Beam is a two-line change once you have a key.