Beam Benchmarks, Specifications & Availability

Beam

Explore Beam from Reflection AI: published specifications, source-linked vendor benchmarks, independent evaluator coverage and recorded pricing when available.

Published specifications

Provider: Reflection AI
Access: Proprietary
License: Not published
Context window: 256K
Total parameters: 501B
Active parameters: 23B
Released: 2026-10-05
Modalities: text
Family: Beam

Announcement · Website

Model notes

Early-access preview; public weights, technical report and model card are not yet released. Reflection plans an Apache 2.0 weights release later in October 2026; the preview license is not published. Official Hugging Face organizations reflection/reflectionai have no public models (checked 2026-10-06). Vendor states 501B total / 23B active parameters and text-only input. Beta API docs specify 262,144-token context and 131,072-token maximum output: https://developers.reflection.ai/models . The launch’s 1M effective midtraining context is not the serving limit. Mandatory reasoning supports low/medium/high/xhigh/max (default medium); launch benchmark efforts and full harness settings are unspecified. No OpenRouter listing or independent Beam evaluation found on 2026-10-06.

Compare Beam with other models →Calculate cost Explore data coverage

Pricing · OpenRouter

No OpenRouter pricing is recorded for this model. Missing rates are not free.

Official / vendor benchmarks

Default headline records. Own-vendor, peer-vendor and third-party provenance remain visible in evidence; configurations may differ.

Official / vendor headline scores; expand evidence for every record
Benchmark / evaluator	Headline score	Evidence
DeepSWE v1.1	44.4%unknown effort	All 1 recorded result & sources 44.4% · raw 44.4 % Headline · unknown effort · Own vendor Source/record date: 2026-10-05 Harness: Reflection launch evaluation; full agent, revision and token-budget settings not published. Beam early-access preview, reported by Reflection. https://reflection.ai/blog/introducing-beam
SWE-bench Pro v2-Hard	77.2%unknown effort	All 1 recorded result & sources 77.2% · raw 77.2 % Headline · unknown effort · Own vendor Source/record date: 2026-10-05 Harness: Reflection launch evaluation; full agent, revision and token-budget settings not published. Beam early-access preview, reported by Reflection. https://reflection.ai/blog/introducing-beam
SWE-bench Pro v1	65.5%unknown effort	All 1 recorded result & sources 65.5% · raw 65.5 % Headline · unknown effort · Own vendor Source/record date: 2026-10-05 Harness: Reflection launch evaluation; full agent, revision and token-budget settings not published. Beam early-access preview, reported by Reflection. https://reflection.ai/blog/introducing-beam
Terminal-Bench 2.1	80.1%unknown effort	All 1 recorded result & sources 80.1% · raw 80.1 % Headline · unknown effort · Own vendor Source/record date: 2026-10-05 Harness: Reflection launch evaluation; full agent, revision and token-budget settings not published. Beam early-access preview, reported by Reflection. https://reflection.ai/blog/introducing-beam
SWE-Atlas Codebase QnA	34.6%unknown effort	All 1 recorded result & sources 34.6% · raw 34.6 % Headline · unknown effort · Own vendor Source/record date: 2026-10-05 Harness: Reflection launch evaluation; full agent, revision and token-budget settings not published. Beam early-access preview, reported by Reflection. https://reflection.ai/blog/introducing-beam
SWE-bench Multilingual	78%unknown effort	All 1 recorded result & sources 78% · raw 78 % Headline · unknown effort · Own vendor Source/record date: 2026-10-05 Harness: Reflection launch evaluation; full agent, revision and token-budget settings not published. Beam early-access preview, reported by Reflection. https://reflection.ai/blog/introducing-beam
SWE-bench Verified	80.9%unknown effort	All 1 recorded result & sources 80.9% · raw 80.9 % Headline · unknown effort · Own vendor Source/record date: 2026-10-05 Harness: Reflection launch evaluation; full agent, revision and token-budget settings not published. Beam early-access preview, reported by Reflection. https://reflection.ai/blog/introducing-beam
AIME 2026	97.8%unknown effort	All 1 recorded result & sources 97.8% · raw 97.8 % Headline · unknown effort · Own vendor Source/record date: 2026-10-05 Harness: Reflection launch evaluation; full agent, revision and token-budget settings not published. Beam early-access preview, reported by Reflection. Language and evaluation protocol not specified. https://reflection.ai/blog/introducing-beam
HLE (text, no tools)	36.2%unknown effort	All 1 recorded result & sources 36.2% · raw 36.2 % Headline · unknown effort · Own vendor Source/record date: 2026-10-05 Harness: Reflection launch evaluation; full agent, revision and token-budget settings not published. Beam early-access preview, reported by Reflection. No tools; corresponding efficiency figure labels HLE text-only. https://reflection.ai/blog/introducing-beam
SciCode	49.7%unknown effort	All 1 recorded result & sources 49.7% · raw 49.7 % Headline · unknown effort · Own vendor Source/record date: 2026-10-05 Harness: Reflection launch evaluation; full agent, revision and token-budget settings not published. Beam early-access preview, reported by Reflection. https://reflection.ai/blog/introducing-beam
CriPT (AA)	16.3%unknown effort	All 1 recorded result & sources 16.3% · raw 16.3 % Headline · unknown effort · Own vendor Source/record date: 2026-10-05 Harness: Reflection launch evaluation; full agent, revision and token-budget settings not published. Beam early-access preview, reported by Reflection. https://reflection.ai/blog/introducing-beam
GPQA Diamond	90.5%unknown effort	All 1 recorded result & sources 90.5% · raw 90.5 % Headline · unknown effort · Own vendor Source/record date: 2026-10-05 Harness: Reflection launch evaluation; full agent, revision and token-budget settings not published. Beam early-access preview, reported by Reflection. https://reflection.ai/blog/introducing-beam
AutomationBench (public split)	37%unknown effort	All 1 recorded result & sources 37% · raw 37 % Headline · unknown effort · Own vendor Source/record date: 2026-10-05 Harness: Reflection launch evaluation; full agent, revision and token-budget settings not published. Beam early-access preview, reported by Reflection. https://reflection.ai/blog/introducing-beam
MCP Atlas	78.7%unknown effort	All 1 recorded result & sources 78.7% · raw 78.7 % Headline · unknown effort · Own vendor Source/record date: 2026-10-05 Harness: Reflection launch evaluation; full agent, revision and token-budget settings not published. Beam early-access preview, reported by Reflection. https://reflection.ai/blog/introducing-beam
τ³-Bench Banking	38%unknown effort	All 1 recorded result & sources 38% · raw 38 % Headline · unknown effort · Own vendor Source/record date: 2026-10-05 Harness: Reflection launch evaluation; full agent, revision and token-budget settings not published. Beam early-access preview, reported by Reflection. https://reflection.ai/blog/introducing-beam
BrowseComp	77.4%unknown effort	All 1 recorded result & sources 77.4% · raw 77.4 % Headline · unknown effort · Own vendor Source/record date: 2026-10-05 Harness: Reflection launch evaluation; full agent, revision and token-budget settings not published. Beam early-access preview, reported by Reflection. Context management enabled; other agent settings unspecified. https://reflection.ai/blog/introducing-beam
DeepSearchQA	80.1%unknown effort	All 1 recorded result & sources 80.1% · raw 80.1 % Headline · unknown effort · Own vendor Source/record date: 2026-10-05 Harness: Reflection launch evaluation; full agent, revision and token-budget settings not published. Beam early-access preview, reported by Reflection. Context management enabled; other agent settings unspecified. https://reflection.ai/blog/introducing-beam
AA-LCR	79.3%unknown effort	All 1 recorded result & sources 79.3% · raw 79.3 % Headline · unknown effort · Own vendor Source/record date: 2026-10-05 Harness: Reflection launch evaluation; full agent, revision and token-budget settings not published. Beam early-access preview, reported by Reflection. https://reflection.ai/blog/introducing-beam
LongBench v2	65.5%unknown effort	All 1 recorded result & sources 65.5% · raw 65.5 % Headline · unknown effort · Own vendor Source/record date: 2026-10-05 Harness: Reflection launch evaluation; full agent, revision and token-budget settings not published. Beam early-access preview, reported by Reflection. https://reflection.ai/blog/introducing-beam
IFBench	79.7%unknown effort	All 1 recorded result & sources 79.7% · raw 79.7 % Headline · unknown effort · Own vendor Source/record date: 2026-10-05 Harness: Reflection launch evaluation; full agent, revision and token-budget settings not published. Beam early-access preview, reported by Reflection. https://reflection.ai/blog/introducing-beam
AA-Omniscience Index (public set)	13unknown effort	All 1 recorded result & sources 13 · raw 13 index Headline · unknown effort · Own vendor Source/record date: 2026-10-05 Harness: Reflection launch evaluation; full agent, revision and token-budget settings not published. Beam early-access preview, reported by Reflection. Public split only, raw index (not percent), evaluated by Reflection; no full/private AA subset implied. https://reflection.ai/blog/introducing-beam
Land or Water grid (180×90 demo)	95.5%unknown effort	All 1 recorded result & sources 95.5% · raw 95.5 % Headline · unknown effort · Own vendor Source/record date: 2026-10-05 Harness: Reflection single-prompt land/water demo, 180×90 grid / 16,200 points; vendor reports 95.5% correct coverage. Grader, repeat count and effort not published. Comparison names Opus 5 and Fable 5 are not the site’s Opus 5.5/Fable 5.1 snapshots; peer results kept only in research notes. https://reflection.ai/blog/introducing-beam

Independent evaluators

Evaluator harnesses are distinct from vendor measurements. Missing coverage is not a failed test.

No independent evaluator results are recorded for this model.

Read how we select and source scores or the comparison guide.