Ollama is offering Reflection’s new Beam model free for a limited time, with access through Ollama Cloud and a zero-data-retention endpoint. The open-weight model is built for coding and long-running agent workflows, with a one-million-token context window.
Ollama Watch analysis
What happened
Ollama says Beam has 501 billion total parameters, with 23 billion active per token. Users can run it through Ollama Cloud from apps or a terminal, without downloading the model or managing local inference. Ollama says its Beam endpoint, provided in collaboration with Reflection, has full zero data retention.
The model is released under Apache 2.0, and Ollama lists launch commands for Claude Code, Codex and OpenCode. The free offer is for a limited time, but no end date is specified.
Why it matters
This gives developers a way to try a very large open-weight model for multi-step coding and tool-using work without first setting up local hardware. The one-million-token context window is aimed at workflows that need to carry substantial material across a long task, rather than just answer a single prompt.
The practical trade-off is that this route runs through Ollama Cloud, not on the user’s own machine. The zero-data-retention statement is useful for evaluating that option, while the temporary free offer makes the cost of trying it unusually clear, at least for now.
Our read
Beam’s appeal is the combination: open weights and a permissive licence, cloud access, and a context window sized for extended work. That is a notable on-ramp for developers, not proof that the model will outperform alternatives on their own tasks. Try it on a representative workflow before handing it the keys to a sprawling codebase. Even a million-token window cannot make a bad answer a good one.
What to watch
- When Ollama’s limited-time free offer ends, and what access costs afterwards.
- Whether Ollama or Reflection publishes more detail on the endpoint’s data-retention arrangement.
- How Beam performs on real coding and long-running agent workflows compared with other available models.
Discussion spark: For coding agents, would you prioritise a huge context window and cloud convenience, or running a smaller model locally with more control over your data?
Sources and evidence
- Source update (16 October 2026, 00:00 UTC)
not affiliated with or endorsed by Ollama