Benchmark Almanac

Beam 1.0

Beam 1.0 places in the top three on 3 benchmarks.

Reflection Released October 5th, 2026 Compare with others
83
Benchmarks with a score
0
Ranked first, out of five or more models
3
In the top three
Reflection Oct 2026

Beam 1.0

Open weights

150.9 #66 of 256
Epoch Capabilities Index
Self-Reported Index 154.0 #67 of 404
  • IFBench 79.7% #10 /80
  • SciCode 49.7% #10 /49
  • AA-LCR (Artificial Analysis Long Context Reasoning) 79.3% #2 /38

Where it ranks

Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.

Knowledge and language

BenchmarkScoreRankSource
AA-LCR (Artificial Analysis Long Context Reasoning) 79.3% #2 of 38 Lab model card
LongBench v2 66.9% #2 of 35 Lab model card
IFBench 79.7% #10 of 80 Lab model card
Humanity's Last Exam 51.2% #37 of 154 Lab model card
LMArena (Elo) 1,435 #84 of 335 Official leaderboard
AA-Omniscience 13.0 #3 of 10 Lab model card
SimpleQA Verified 41.8% #13 of 22 Lab model card

Math and science

BenchmarkScoreRankSource
GPQA Diamond 90.5% #29 of 302 Lab model card
AIME 2026 97.8% #4 of 38 Lab model card
CritPt 16.3% #6 of 13 Lab model card
ARC-AGI-2 23.5% #51 of 98 Official leaderboard
ARC-AGI 74% #52 of 95 Official leaderboard
MysteryMechanism 22.1% #23 of 31 Official leaderboard
Riemann-bench 37.6% #16 of 20 Official leaderboard
ProofBench v1.1 18% #43 of 52 Official leaderboard
BioMysteryBench 42.6% #30 of 30 Official leaderboard

Coding

BenchmarkScoreRankSource
SWE-bench Verified 80.9% #10 of 155 Lab model card
LiveCodeBench 88.3% #5 of 54 Lab model card
SWE-Bench Pro 65.5% #13 of 81 Lab model card
SciCode 49.7% #10 of 49 Lab model card
SWE-bench Multilingual 78% #21 of 73 Lab model card
DesignArena UI Component 1,259 #46 of 139 Official leaderboard
DesignArena Website 1,245 #60 of 151 Official leaderboard
DesignArena Code 1,227 #64 of 143 Official leaderboard
Vibe Code Bench 66.4% #54 of 120 Official leaderboard
ALE-Bench 1,493 #57 of 126 Official leaderboard
DesignArena Game Dev 1,186 #73 of 143 Official leaderboard
Code Arena 1,437 #73 of 128 Official leaderboard
DesignArena 3D Design 1,123 #82 of 136 Official leaderboard
Code Migration v1.1 22.9% #26 of 30 Official leaderboard
DeepSWE 44.4% #51 of 56 Lab model card
IOI v2 (Vals) 22.8% #41 of 45 Official leaderboard
OWE-Bench 3.30% #13 of 14 Official leaderboard
FrontierCode 2.0 3.41% #31 of 32 Official leaderboard
SWE Atlas 34.6% #14 of 14 Lab model card
SWE-Bench Pro V2 Hard 77.2% #3 of 3 Lab model card

Agents and tools

BenchmarkScoreRankSource
τ³-bench Banking 38% #5 of 22 Lab model card
MCP Atlas 78.7% #17 of 46 Lab model card
RuneBench 5.26 #29 of 76 Official leaderboard
BrowseComp 77.4% #36 of 81 Lab model card
Terminal-Bench 2.1 80.1% #29 of 56 Lab model card
AutomationBench 37% #26 of 47 Lab model card
DeepSearchQA 80.1% #15 of 23 Lab model card
Agent Arena -5.91% #44 of 56 Official leaderboard
Terminal-Bench 4.0 6.36% #36 of 40 Official leaderboard
Terminal-Bench 4.1 6.36% #39 of 43 Official leaderboard

Professional work

BenchmarkScoreRankSource
EnterpriseBench: CoreCraft 49.2% #24 of 51 Official leaderboard
APEX-Agents 48.8% #36 of 61 Collected by Epoch AI, source not named
Legal Research Bench 28.9% #60 of 93 Official leaderboard
Tuesday Work Index 40.3 #33 of 44 Official leaderboard
DAYJOB: Finance 1.50% #31 of 39 Official leaderboard
DAYJOB: Healthcare 1.20% #30 of 37 Official leaderboard
Harvey LAB v1.1 (HLAB-Vals-1.1) 4.58% #32 of 38 Official leaderboard
Tax Agent Bench 8.81% #72 of 84 Official leaderboard
Excel Modeling Benchmark v1.1 64.7% #36 of 42 Official leaderboard
Vals Legal Index 61.8% #34 of 38 Official leaderboard
Vals Index v2.2 43.4% #34 of 37 Official leaderboard
Vals Finance Agent v2.1 51.8% #37 of 40 Official leaderboard
Vals Finance Index 46.5% #37 of 40 Official leaderboard

Safety and security

BenchmarkScoreRankSource
Human Pathogen Capabilities Test 60.7% #22 of 86 Official leaderboard
ABC Bench Liquid-Handling Robot 96.7% #13 of 34 Official leaderboard
Human Pathogen Capabilities Test v2 (HPCT v2) 59.6% #30 of 71 Official leaderboard
World-Class Bio 48.8% #33 of 74 Official leaderboard
Molecular Biology Capabilities Test 49.2% #44 of 90 Official leaderboard
BioTIER Refuse 61.3% #35 of 69 Official leaderboard
Virology Capabilities Test (VCT) 52% #17 of 31 Lab model card
Agentic BAIM-LLM Evaluation (ABLE) 50.5% #11 of 18 Official leaderboard
ABC Bench Screening Evasion 56% #11 of 17 Official leaderboard
TxBench-Antibody-Discovery 34% #22 of 30 Official leaderboard
ABC Bench Fragment Design 77.5% #20 of 27 Official leaderboard
BioTIER Permit 98.5% #53 of 69 Official leaderboard
BioSecBench-Refusal 3.40% #19 of 22 Official leaderboard
TxBench-Oligo-Discovery 27.4% #26 of 30 Official leaderboard
TxBench-Preclinical-Pharmacology 39.7% #26 of 30 Official leaderboard
OncologyBench 1.33% #29 of 30 Official leaderboard
CyScenarioBench 0% #17 of 17 Lab model card
WMDP Chemistry 62.4% #7 of 7 Lab model card
Atomic Challenges 57.5% #3 of 3 Lab model card
Iris Safety Evaluations 88.9% #1 of 1 Lab model card
MirrorDojo 3.10% #1 of 1 Lab model card
MirrorDyn 3.60% #1 of 1 Lab model card

Other

BenchmarkScoreRankSource
Political Science LLM Benchmark 66.2% #26 of 42 Official leaderboard
Vals Science Index 22% #25 of 25 Official leaderboard

Its lab's numbers put it at 154.0, 3.1 points above its ECI

The Self-Reported Index fits one score to the 21 capability results Reflection reported for this model, on the scale of the Epoch Capabilities Index. Labs choose which benchmarks to show and how to run them, so read it as the lab's claim, not an independent measurement. The 90% interval runs from 152.0 to 156.0. How we compute it.

Its lab's numbers next to 6 leaderboards

Where the benchmark's own leaderboard, or Epoch AI's run, also has this model. Gaps mostly come from different harnesses, effort levels and task subsets. Every pair.

Benchmark Lab says Board says Gap
World-Class Bio 37.9% World-Class Biology accuracy 48.8% Leaderboard, max −11 pts
ABC Bench Fragment Design 82.1% ABC-Bench Fragment Design, 23 of 28 completed attempts passed, Pass@1 77.5% Leaderboard, max +4.6 pts
BioTIER Refuse 64.6% BioTIER refuse split accuracy 61.3% Leaderboard, max +3.3 pts
BioTIER Permit 97.2% BioTIER permit split accuracy 98.5% Leaderboard, max −1.3 pts
Molecular Biology Capabilities Test 48.5% MBCT exact-match accuracy 49.2% Leaderboard, max −0.7 pts
Human Pathogen Capabilities Test 61% HPCT accuracy 60.7% Leaderboard, max +0.3 pts

Reflection's models over time

Reflection's models in color, against every other model on the Epoch Capabilities Index, by release date.

50 100 150 2023 2024 2025 2026 2027 Beam 1.0 50 100 150 2023 2024 2025 2026 2027 Beam 1.0

Source: Epoch Capabilities Index, Epoch AI (CC BY 4.0)

Its launch post

The benchmarks the lab chose to headline, with the scores it gave.

Introducing Beam (Reflection blog) Oct 5, 2026

Its system card

Every result in the longer documents the lab published with it. 15 of them aren't in the launch post.