Identity
- Status
- preview
- Developer ID
Beam-501B-A23B- Released
- Oct 5, 2026
- Version
- Beam
- License
- Unknown
Reflection AIBeam
Reflection AI's text-only sparse mixture-of-experts reasoning model. Its 262,144-token API context and 131,072-token output limit are documented for a selective beta; weights and public pricing have not been released.
- Token context
- 262K tokens
- Inputs
- Text
- Outputs
- Text
- Released
- Oct 5, 2026
Providers & Pricing
1 current hosted offerings
No Directly Sourced Provider Price Yet
API pricing is provider-specific and stays unknown until a serving provider publishes an exact offering for this model version.
At a Glance
Model Facts
Capacity
- Token context
- 262K tokens
- Maximum output
- 131K tokens
- Total parameters
- 501B
- Active parameters
- 23B
- Knowledge cutoff
- Jun 30, 2026
Interface and Access
- Inputs
- Text
- Outputs
- Text
- API available
- Yes
- Open weights
- No
- Self-hostable
- No
- Reasoning
- Yes
- Vision
- No
- Tool calling
- Yes
- Capabilities
- chat, generation, reasoning, structured outputs, tools
Technical Specifications
- Architecture design
- Sparse mixture-of-experts
- Reasoning efforts
- low, medium, high, xhigh, max; reasoning required
- API access
- Selective waitlist beta
- Weights
- Announced for later in October 2026; not yet published
This page represents one developer model product. Dated API snapshots, serving endpoints, pricing, regions, quantizations, and service tiers attach as versioned aliases or provider details and never create duplicate public model records.
Version Lineage
PredecessorUnknown
SuccessorsUnknown
AliasesBeam-501B-A23B
Explicit Unknowns
- Tokenizer
- Provider pricing
Primary Evidence
Sources and Observation Date
Comparable Peers
Related Model Comparisons
No Eligible Peers YetComparisons require a shared sourced modality, workload, and enough material differences.
Questions
Beam FAQs
What is Beam?+
Reflection AI's text-only sparse mixture-of-experts reasoning model. Its 262,144-token API context and 131,072-token output limit are documented for a selective beta; weights and public pricing have not been released. It is developed by Reflection AI and its current sourced lifecycle status is preview.
What inputs and outputs does Beam support?+
Beam's current record lists Text as input and Text as output.
How much context does Beam support?+
The current record lists a 262K-token context window and a maximum output of 131K tokens.
What technical specifications are published for Beam?+
Beam's developer-published specifications include architecture design: Sparse mixture-of-experts; reasoning efforts: low, medium, high, xhigh, max; reasoning required; api access: Selective waitlist beta; weights: Announced for later in October 2026; not yet published.
How large is Beam?+
The current record lists 501B total parameters and 23B active parameters. Its architecture is Sparse mixture-of-experts.
What capabilities and tasks does Beam support?+
Beam's recorded capabilities are chat, generation, reasoning, structured outputs, and tools. Its supported tasks are coding, reasoning, and text.
Is Beam available through an API?+
Beam is marked as API-available. The directly sourced provider records currently include Reflection AI.
How much does Beam cost through an API?+
No directly sourced token price is currently attached to Beam. Missing prices remain unknown rather than being inferred from another model or provider.
Who can use Beam today, and are its weights downloadable?+
Reflection AI documents an API route, but access is opening gradually to selected waitlist users. Its October 5 announcement says weights, a model card, and a technical report are planned for later in the month; they are not yet published.
Can Beam be self-hosted?+
Beam's weights are not yet published, so self-hosting is not currently available. Reflection AI has announced a later weight release, but no operative license is recorded yet.