Reflection's open-weight Beam model is free on Ollama for a limited time
Reflection's first open-weight model, Beam, is now generally available, and Ollama is offering it for free for a limited time through Ollama Cloud. According to Ollama, its Beam endpoint runs with full zero data retention (ZDR).
- Reflection's first open-weight model, Beam, is generally available and free on Ollama for a limited time.
- Beam is a mixture-of-experts model with approximately 501 billion total and 23 billion active parameters per token.
- It offers a 1 million token context window and is released under Apache 2.0.
- Ollama's Beam endpoint runs with full zero data retention, according to Ollama.
- Run it with ollama run beam:cloud or launch it with Claude Code, Codex, or OpenCode.
What Beam is and how it is built
Beam is a mixture-of-experts model with approximately 501 billion total parameters and 23 billion active parameters per token, according to Ollama's announcement. It is designed for software engineering, reasoning, and long-horizon agentic tasks.
The model is aimed at workflows that go beyond a single response: writing and maintaining code, using tools across multiple steps, and working through complex tasks over time. Beam uses a text-in, text-out design, which Ollama says focuses on workloads where code, tool calls, logs, documents, and structured outputs are central.
To support long-running workflows, Beam offers a 1 million token context window. It is released under the Apache 2.0 license, which gives developers the freedom to inspect, customize, and build on the model.
Free access through Ollama Cloud
Ollama says it is a day-zero partner for the launch and is rolling Beam out for free, though only for a limited time. The source does not say how long the free period lasts or what pricing will follow.
Beam is available through Ollama Cloud, where users can run supported models from their apps or terminal without downloading models or managing local inference. In collaboration with Reflection, Ollama's Beam endpoint runs with full zero data retention.
Ollama lists several intended use cases:
- Coding agents: solving software engineering tasks, working through codebases, and helping write and maintain code.
- Long-running agents: executing multi-step workflows, using tools, and working toward a goal over an extended sequence of actions.
- Math and science reasoning: working through complex mathematical and scientific problems.
- Custom models: fine-tuning and customizing an open-weight model for specific applications.
How to get started
Users need to download the latest version of Ollama. The basic command to run the model is ollama run beam:cloud.
Ollama also provides launch commands for coding tools, each using the same beam:cloud model name:
- Claude Code: ollama launch claude --model beam:cloud
- Codex: ollama launch codex --model beam:cloud
- OpenCode: ollama launch opencode --model beam:cloud
Ollama points to a Beam model page, Reflection's Beam announcement, and the Reflection Beam model card for more information. The announcement mentions a benchmarks section, but the source text provides no benchmark figures.
What it means for developers
For developers, the offer lowers the barrier to trying a large open-weight model aimed at agentic coding. Because the model runs in Ollama Cloud, there is no need to manage local inference, and the Apache 2.0 license leaves room to customize and build on it. The zero data retention setup on Ollama's endpoint is another point the company highlights. Since the free period is limited, teams evaluating Beam have a reason to test it soon, though the source gives no end date.
