AI Agent StoreReflection · Early access beta
Reflection Beam
A text model for coding and tool-based work, currently behind beta access. Its API limits and five effort settings are documented; comparable independent results and public token prices are not verified.
Released 2026-10-05 · First reviewed 2026-10-07 · Checked 2026-10-15.
We still need comparable independent tests. You can compare the confirmed features and limits below, but we cannot yet give this version an Agent Fit or Practical Value rating. Test it on your own tasks before switching; results for another version or thinking setting may not apply.
What you can use today
Token prices measure the small pieces of text the model reads or produces. Capacity is the amount the service accepts in a request; check the limits below for any fallback to another model. A dash means the rate is unverified.
Exact provider API ID: Beam-1.0-501B-A23B. Beam supports tools and structured answers, with low, medium, high, xhigh and max thinking. None and minimal are unsupported. Reflection documents defaults that can differ by organization; check your account’s model response. The documented API name is now Beam-1.0-501B-A23B; this documentation change alone does not establish a new checkpoint or improved quality.
Available thinking settings: low, medium, high, xhigh, max. We do not yet have comparable independent results for these settings. The default applied when a request omits effort is unverified.
New accounts join a waitlist and can create keys after access is enabled. Weights and developer artifacts are promised later in October; they are not available simply because the model is announced.
The beta lists 262,144 combined input/output tokens and up to 131,072 output tokens. Your plan may allow less. Requests that ask for token probabilities (logprobs) allow at most 8,192 output tokens, including thinking. Set an output budget that fits alongside your prompt.
Public per-token prices are not verified. Check your account contract; neither beta access nor planned open weights establishes free inference. Your bill can also include tools, retries and charged thinking tokens. Check the offer used by your application.
Provider release and specifications ↗
Check availability and plans ↗
Test it before switching
Give each model the same examples, tools and definition of success. Record correct outcomes, response time, retries and total cost. A cheaper token price does not guarantee a cheaper completed task.
Compare caching and long-context costs ↗