# ReflectionBeam.org

> Free tools for building with Reflection AI's Beam: a GPU memory planner for self-hosting, a token and cost calculator, an API code generator and ready-made coding agent configs.

ReflectionBeam.org is an independent set of developer tools for Beam, the open-weight model from Reflection AI. Facts come from Reflection's developer documentation and are checked regularly.

## Tools

- [GPU memory planner](https://reflectionbeam.org/vram): How much VRAM Beam needs at each precision and context length, and how many GPUs that is.
- [Token and cost calculator](https://reflectionbeam.org/cost): Estimate tokens per request with reasoning, daily volume against rate limits, and self-hosting cost.
- [API code generator](https://reflectionbeam.org/code): Working curl, Python and TypeScript for streaming, tool calls, JSON output and multi-turn chats.
- [Coding agent configs](https://reflectionbeam.org/agents): Copy-ready setup for Mirror CLI, Pi, OpenCode and Hermes Agent with Beam.

## Key facts

- Model ID: Beam-501B-A23B
- OpenAI-compatible base URL: https://api.reflection.ai/openai/v1 (Chat Completions and Models only)
- Parameters: 501B total, 23B active (Mixture-of-Experts), 52 layers
- Context: 262,144 tokens; max output 131,072 tokens
- reasoning_effort: low, medium, high, xhigh, max (default medium; reasoning cannot be disabled)
- Earlier assistant messages must be sent back with reasoning_content, or the API returns 400 missing_required_parameter

## Site

- [About](https://reflectionbeam.org/about)
- [Privacy](https://reflectionbeam.org/privacy)

## Optional

- Every page is also available as Markdown: request it with the header `Accept: text/markdown`.
- Languages: English at the root; ru, es, fr, de, pt, it, nl, pl, tr, zh, ar, id, ja, ko, fil, th, vi under /<lang>, for example https://reflectionbeam.org/ru/vram.
- [Reflection developer docs](https://developers.reflection.ai/)
- [ReflectionBeam.com, the Beam guide](https://reflectionbeam.com)
- [BeamAI.chat, a free chat](https://beamai.chat)