evals.report
BenchmarksLabsCompareRun guidesIn the wild

Beam 1.0

Reflection AI · Beam. Released Oct 16, 2026.

Beam 1.0 is a model from Reflection AI in the Beam family, released Oct 16, 2026. evals.report tracks 14 reported Beam 1.0 benchmark scores across ARC-AGI-2, ARC-AGI-1, FrontierCode 2.0, AA-Omniscience v1.1: Knowledge and Hallucination Benchmark, AA-Omniscience: Knowledge and Hallucination Benchmark, Terminal-Bench 2.1, SWE-bench Pro, SWE-bench Multilingual, and 6 more — each shown with its benchmark, metric, source status, and date, and never combined into a single ranking.

Open14 results

Benchmark results 14

Compare this model
BenchmarkCategoryScoreMetricStatusDate
ARC-AGI-2Reasoning23.47%accuracyOfficialOct 16, 2026Details
ARC-AGI-1Reasoning74%accuracyOfficialOct 16, 2026Details
FrontierCode 2.0Coding3.4%weighted score (Main)OfficialOct 16, 2026Details
AA-Omniscience v1.1: Knowledge and Hallucination BenchmarkReasoning2AA-Omniscience Net ScoreOfficialOct 16, 2026Details
AA-Omniscience: Knowledge and Hallucination BenchmarkReasoning2AA-Omniscience IndexOfficialOct 16, 2026Details
Terminal-Bench 2.1Agents80.1%task successVerifiedOct 16, 2026Details
SWE-bench ProCoding65.5%% resolvedVerifiedOct 16, 2026Details
SWE-bench MultilingualCoding78.0%% resolvedVerifiedOct 16, 2026Details
AIME 2026Reasoning97.8%accuracyVerifiedOct 16, 2026Details
Humanity's Last ExamReasoning36.2%accuracyVerifiedOct 16, 2026Details
GPQA DiamondReasoning90.5%accuracyVerifiedOct 16, 2026Details
BrowseCompAgents77.4%accuracyVerifiedOct 16, 2026Details
IFBenchReasoning79.7%accuracyVerifiedOct 16, 2026Details
SciCodeCoding49.7%accuracyVerifiedOct 16, 2026Details