Beam Models by Reflection AI
Reflection AI releases · 1 in the last 12 monthsChangelog →
1 model2026Up to 1M ctxFrom $1.30/1M input
Last refreshed 2026-10-16. Next refresh: weekly.
Details
ResearcherReflection AI
LicenseApache 2.0OSI-approved
Commercial useCommercial use: permitted
Models1
Released2026
Max context1M
Capabilities
ReasoningAll models
JSON / Tool useAll models
Structured OutputsAll models
About
Beam is Reflection AI's family of open-weight mixture-of-experts language models for coding, reasoning, and agentic work, trained from scratch. Beam 1.0 (501B total, 23B active parameters) was announced on 5 October 2026 as a waitlisted API preview. Reflection released the weights on Hugging Face under Apache 2.0 on 16 October 2026 and opened self-serve API keys.
Decision facts
Current Variants
Use-when guidance is based on each model's tracked capabilities, context window, release date, and replacement status.
1 in view
BeamCurrent
Use when the workload needs 1M context, reasoning, and JSON / Tool use.
2026-101M contextreasoningJSON / Tool use
| Model | Use when | Released | Signals | Status |
|---|---|---|---|---|
| Beam | Use when the workload needs 1M context, reasoning, and JSON / Tool use. | 2026-10 | 1M contextreasoningJSON / Tool use | Current |
Release Timeline
1 release group2026-10
1 current
Beam
Current1M contextreasoningJSON / Tool use
Specifications(1 models)
| Model | Released | Context | Parameters | Reasoning | JSON / Tool use | Structured Outputs |
|---|---|---|---|---|---|---|
| Beam | 2026-10 | 1M | 501B (23B activated) | Yes | Yes | Yes |
Available From(1 provider)
Pricing
| Model | Provider | Input / 1M | Output / 1M | Type |
|---|---|---|---|---|
| Beam | Vercel AI Gateway | $1.3 | $4.3 | Serverless |