# Beam 1.0 501B-A23B NVFP4 Benchmark Scores & Performance

> Beam 1.0 501B-A23B NVFP4 by Reflection AI has 20 source-displayable benchmark rows but no public overall score. It remains unranked.

Canonical page: https://benchlm.ai/models/reflection-beam-nvfp4

Last updated: October 16, 2026

## Model Details

| Property | Value |
|----------|-------|
| Creator | Reflection AI |
| Source Type | Open Weight |
| Reasoning Type | Reasoning |
| Context Window | 1M |
| Official model card | [Reflection NVFP4 model card](https://huggingface.co/reflection/Beam-1.0-501B-A23B-NVFP4) |
| Overall Score | Not computed (unranked) |
| Overall Rank | Unranked |

## Family & Coverage

- Family: Beam
- Variant: nvfp4 (NVFP4)
- Benchmarks covered: 20 of 639
- Sibling models: [Beam 1.0 501B-A23B](/models/reflection-beam), [Beam 1.0 501B-A23B FP8](/models/reflection-beam-fp8)
- Coverage note: Published evidence is shown below; no public overall score is assigned.

## Agentic Benchmarks

| Benchmark | Score | Source |
|-----------|-------|--------|
| [τ³-bench results](/benchmarks/tau3-bench) | 35.4% | [Reflection Beam 1.0 NVFP4 model card](https://huggingface.co/reflection/Beam-1.0-501B-A23B-NVFP4) |
| [HLE w/ tools](/benchmarks/hlewithtools) | 51.1% | [Reflection Beam 1.0 NVFP4 model card](https://huggingface.co/reflection/Beam-1.0-501B-A23B-NVFP4) |
| [Terminal-Bench 2.1](/benchmarks/terminalbench21) | 78.4% | [Reflection Beam 1.0 NVFP4 model card](https://huggingface.co/reflection/Beam-1.0-501B-A23B-NVFP4) |
| [MCP Atlas](/benchmarks/mcpatlas) | 80.5% | [Reflection Beam 1.0 NVFP4 model card](https://huggingface.co/reflection/Beam-1.0-501B-A23B-NVFP4) |
| [BrowseComp](/benchmarks/browsecomp) | 75.6% | [Reflection Beam 1.0 NVFP4 model card](https://huggingface.co/reflection/Beam-1.0-501B-A23B-NVFP4) |
| [DeepSearchQA](/benchmarks/deepsearchqa) | 81.1% | [Reflection Beam 1.0 NVFP4 model card](https://huggingface.co/reflection/Beam-1.0-501B-A23B-NVFP4) |

## Coding Benchmarks

| Benchmark | Score | Source |
|-----------|-------|--------|
| [DeepSWE](/benchmarks/deepswe) | 42.7% | [Reflection Beam 1.0 NVFP4 model card](https://huggingface.co/reflection/Beam-1.0-501B-A23B-NVFP4) |
| [SWE-bench Pro](/benchmarks/swe-bench-pro) | 65.6% | [Reflection Beam 1.0 NVFP4 model card](https://huggingface.co/reflection/Beam-1.0-501B-A23B-NVFP4) |
| [Terminal-Bench 2.1](/benchmarks/terminalbench21) | 78.4% | [Reflection Beam 1.0 NVFP4 model card](https://huggingface.co/reflection/Beam-1.0-501B-A23B-NVFP4) |
| [SWE-Atlas Codebase QnA](/benchmarks/swe-atlas-codebase-qna) | 31.4% | [Reflection Beam 1.0 NVFP4 model card](https://huggingface.co/reflection/Beam-1.0-501B-A23B-NVFP4) |
| [SWE Multilingual](/benchmarks/swe-bench-multilingual) | 77.5% | [Reflection Beam 1.0 NVFP4 model card](https://huggingface.co/reflection/Beam-1.0-501B-A23B-NVFP4) |
| [SWE-bench Verified](/benchmarks/swe-bench-verified) | 80.7% | [Reflection Beam 1.0 NVFP4 model card](https://huggingface.co/reflection/Beam-1.0-501B-A23B-NVFP4) |
| [SciCode](/benchmarks/scicode) | 46.9% | [Reflection Beam 1.0 NVFP4 model card](https://huggingface.co/reflection/Beam-1.0-501B-A23B-NVFP4) |

## Reasoning Benchmarks

| Benchmark | Score | Source |
|-----------|-------|--------|
| [AA-LCR](/benchmarks/lcr) | 78.7% | [Reflection Beam 1.0 NVFP4 model card](https://huggingface.co/reflection/Beam-1.0-501B-A23B-NVFP4) |
| [CritPt](/benchmarks/critpt) | 15.4% | [Reflection Beam 1.0 NVFP4 model card](https://huggingface.co/reflection/Beam-1.0-501B-A23B-NVFP4) |
| [LongBench v2](/benchmarks/longbench-v2) | 67.2% | [Reflection Beam 1.0 NVFP4 model card](https://huggingface.co/reflection/Beam-1.0-501B-A23B-NVFP4) |

## Knowledge Benchmarks

| Benchmark | Score | Source |
|-----------|-------|--------|
| [GPQA-D](/benchmarks/gpqa-diamond) | 89.3% | [Reflection Beam 1.0 NVFP4 model card](https://huggingface.co/reflection/Beam-1.0-501B-A23B-NVFP4) |
| [HLE w/o tools](/benchmarks/hlenotools) | 34.8% | [Reflection Beam 1.0 NVFP4 model card](https://huggingface.co/reflection/Beam-1.0-501B-A23B-NVFP4) |

## Instruction Following Benchmarks

| Benchmark | Score | Source |
|-----------|-------|--------|
| [IFBench](/benchmarks/ifbench) | 79.2% | [Reflection Beam 1.0 NVFP4 model card](https://huggingface.co/reflection/Beam-1.0-501B-A23B-NVFP4) |

## Mathematics Benchmarks

| Benchmark | Score | Source |
|-----------|-------|--------|
| [AIME26](/benchmarks/aime2026) | 97.5% | [Reflection Beam 1.0 NVFP4 model card](https://huggingface.co/reflection/Beam-1.0-501B-A23B-NVFP4) |

## Other Reflection AI Models

- [Beam 1.0 501B-A23B](/models/reflection-beam) - Score: not computed
- [Beam 1.0 501B-A23B FP8](/models/reflection-beam-fp8) - Score: not computed