Beam
Released
2026-10-05
Last refreshed
2026-10-16
Status
Researched today
Open sourceCommercial use: permittedCodingRAGAgentsLong contextClassificationJSON / Tool useAgentCoding
Reflection AI releases · 1 in the last 12 monthsChangelog →
Specifications
- Family
- Beam
- Released
- 2026-10-05
- Context
- 1M
- Max output
- 128K
- Parameters
- 501B (23B activated)
- Architecture
- Mixture of Experts
- Knowledge cutoff
- 2026-06
- Specialization
- general
- Openness
- Open source
- License
- Apache 2.0OSI-approvedCommercial use: permitted
- Weights
- Available
- Code
- Unknown
- Training
- Multi-stage
Created by
Pricing
Output / 1M
$4.30
Input / 1M
$1.30
Cheapest of 1 route · Vercel AI Gateway · cache read $0.130
Links
About
Beam is an open-weight sparse mixture-of-experts language model with 501B total and 23B active parameters, built for coding, reasoning, and long-horizon agentic work and trained from scratch. It is text-only, with a 1M-token context window, up to 128K output tokens, five reasoning effort levels (low to max, default medium), tool calling, and structured outputs, and a June 30, 2026 knowledge cutoff.
Top use-case fit: coding, agents, and build tasks
Coding
Q/$ DAgents
Q/$ CProviders(1)
Compare API pricing across 1 tracked provider route for input and output tokens, batch, and cached reads when available.
| Provider | Input / 1M | Output / 1M | Cache | Route |
|---|---|---|---|---|
| Vercel AI Gateway | $1.30 | $4.30 | read $0.130 | Serverless |
Capabilities
ReasoningJSON / Tool useStructured Outputs
Benchmark peer barsfor Coding
SWE-bench ProRank 5 of 51
SWE-bench VerifiedRank 10 of 99
Benchmark scores(12)
Scores are benchmark-specific and are direction-aware: the same numeric gap can mean very different outcomes across suites. Use the leaderboard context and this model's provider route to decide whether the winning margin is meaningful for your workload.
| Benchmark | Score | Version | Evaluation | Source |
|---|---|---|---|---|
| SWE-bench Verified | 81.6 | SWE-bench VerifiedObserved 2026-10-16 | — | Source |
| SWE-bench Multilingual | 78.3 | SWE-bench MultilingualObserved 2026-10-16 | — | Source |
| SWE-bench Pro | 66.1 | SWE Bench Pro v1Observed 2026-10-16 | — | Source |
| Terminal-Bench 2.1 | 80.1 | Terminal Bench v2.1Observed 2026-10-16 | — | Source |
| DeepSWE 1.1 | 43.3 | DeepSWE v1.1Observed 2026-10-16 | — | Source |
| AIME 2026 | 97.7 | AIME 2026Observed 2026-10-16 | — | Source |
| Humanity's Last Exam — No Tools | 36.0 | HLE no toolsObserved 2026-10-16 | — | Source |
| Google-Proof Q&A | 90.6 | diamondObserved 2026-10-16 | — | Source |
| MCP-Atlas | 80.8 | MCP AtlasObserved 2026-10-16 | — | Source |
| BrowseComp | 76.8 | BrowseComp w/ context managementObserved 2026-10-16 | — | Source |
| DeepSearchQA | 80.5 | DeepSearchQA w/context managementObserved 2026-10-16 | — | Source |
| AutomationBench | 39.3 | AutomationBench publicObserved 2026-10-16 | — | Source |
Migration checks
No linked migration route is available for this model yet.
API versions
Beam-1.0-501B-A23Breflection/Beam-1.0-501B-A23Breflection/Beam-1.0-501B-A23B-FP8reflection/Beam-1.0-501B-A23B-NVFP4Beam-501B-A23Breflection/beamCreated by
Pricing
Output / 1M
$4.30
Input / 1M
$1.30
Cheapest of 1 route · Vercel AI Gateway · cache read $0.130
Links