Reflection's open Beam model is free on OpenCode Zen and Go with 1M context
OpenCode added Beam Free, the free version of Reflection's open-source Beam 1.0, to both Zen and Go. Input, output, and cache reads cost nothing, and it doesn't use Go plan quota. It offers about 1M tokens of context and up to about 131K output tokens, with reasoning and tool calling, and no end date has been announced.
The briefing
OpenCode has added Beam Free (model ID beam-free), the free version of Reflection's newly open-sourced Beam 1.0. Anyone with a Zen account or a Go subscription can use it at no cost.
What you can use
Input, output, and cache reads on Beam Free are all billed at $0, and calls from Go don't count against your plan quota. It supports about 1M tokens of context and up to about 131K output tokens, with reasoning, tool calling, and structured JSON output. Reasoning effort can be set to low, medium, high, xhigh, or max.
OpenCode hasn't announced an end date or usage cap for the free period. Requests made during the free period may be used to improve the model, so switch to a paid model for sensitive code.
Specs
Beam 1.0 is a mixture-of-experts model with 501B total parameters and 23B active per token, built for coding, reasoning, and agentic work, with open weights under Apache 2.0. It takes text input only, and its knowledge cutoff is around June 2026.
Beam 1.0 (Free) is also available on Vercel AI Gateway if you'd rather call it there.
How to use it
- Open OpenCode, sign in to Zen or Go, and switch to Beam Free with
/models, or set the model toopencode/beam-freein your config. - For the API, use the Zen endpoint
https://opencode.ai/zen/v1/chat/completionswith modelbeam-free.
More details: OpenCode Zen docs · OpenCode models on models.dev · Beam launch post on Hugging Face.