# GPT-5.4 vs Beam 1.0 501B-A23B FP8

> Side-by-side benchmark comparison for GPT-5.4 and Beam 1.0 501B-A23B FP8 across agentic, coding, multimodal, knowledge, reasoning, multilingual, and math tasks.

- Canonical page: https://benchlm.ai/compare/gpt-5-4-vs-reflection-beam-fp8
- Last updated: October 16, 2026

- Shared sourced benchmarks: 6
- HTML indexing: indexable
- Ranking lane: BenchAlign v5.8

## Quick Verdict

Pick Beam 1.0 501B-A23B FP8 if you want the stronger benchmark profile. GPT-5.4 only makes more sense when its price, context window, or workload-specific category wins matter more than the overall score.

## Summary

- Beam 1.0 501B-A23B FP8 leads overall 71 to 66.69.
- The clearest category separation is in reasoning, where the averages are 64.2 for GPT-5.4 and 85.8 for Beam 1.0 501B-A23B FP8.
- The biggest single benchmark swing is MCP Atlas in Agentic, with scores of 70.6% and 81.5%.
- Beam 1.0 501B-A23B FP8 is the cheaper option on output tokens, which matters if you expect large responses or heavy interactive use.
- GPT-5.4 also has the larger context window at 1.05M.

## Model Snapshot

| Property | GPT-5.4 | Beam 1.0 501B-A23B FP8 |
|----------|----------|----------|
| Creator | OpenAI | Reflection AI |
| Type | Proprietary | Open Weight |
| Reasoning | Reasoning | Reasoning |
| Context | 1.05M | 1M |
| Overall Score | 66.69 | null |
| Benchmarks Covered | 42 | 19 |
| Pricing (input/output) | $2.50 / $15.00 | Pricing unavailable / Pricing unavailable |

## Category Breakdown

### Agentic

- Winner: Not comparable
- GPT-5.4 public-lane score: 52.1 (Supported · #44/128)
- Beam 1.0 501B-A23B FP8 public-lane score: Coming soon

| Benchmark | GPT-5.4 | Beam 1.0 501B-A23B FP8 | Winner |
|-----------|-----------|-----------|--------|
| Terminal-Bench 2.0 | 75.1% | Coming soon | Coming soon |
| CyberGym | 79.0% | Coming soon | Coming soon |
| BrowseComp | 82.7% | 76.7% | GPT-5.4 |
| OSWorld-Verified | 75% | Coming soon | Coming soon |
| MCP Atlas | 70.6% | 81.5% | Beam 1.0 501B-A23B FP8 |
| Toolathlon | 54.6% | Coming soon | Coming soon |
| τ²-bench results | 98.9% | Coming soon | Coming soon |
| Claw-Eval | 60.3% | Coming soon | Coming soon |
| DeepSearchQA | 73.6% | 79.6% | Beam 1.0 501B-A23B FP8 |
| Gert Labs | 64.89% | Coming soon | Coming soon |
| ResearchClawBench | 15.3% | Coming soon | Coming soon |
| JobBench | 38.9% | Coming soon | Coming soon |
| ExploitGym | 6.0% | Coming soon | Coming soon |
| ApprenticeBench | 11% | Coming soon | Coming soon |
| MazeBench | 0% | Coming soon | Coming soon |
| τ³-bench results | Coming soon | 36.8% | Coming soon |
| HLE w/ tools | Coming soon | 51.1% | Coming soon |
| Terminal-Bench 2.1 | Coming soon | 78.9% | Coming soon |

### Coding

- Winner: Not comparable
- GPT-5.4 public-lane score: 47.7 (Supported · #48/150)
- Beam 1.0 501B-A23B FP8 public-lane score: Coming soon

| Benchmark | GPT-5.4 | Beam 1.0 501B-A23B FP8 | Winner |
|-----------|-----------|-----------|--------|
| LiveCodeBench Pro | 87.5% | Coming soon | Coming soon |
| SWE-bench Pro | 57.7% | 65.9% | Beam 1.0 501B-A23B FP8 |
| React Native Evals | 85.3% | Coming soon | Coming soon |
| Terminal-Bench 2.0 | 75.1% | Coming soon | Coming soon |
| Vibe Code Bench | 67.42% | Coming soon | Coming soon |
| PostTrainBench v1.1 | 19.0% | Coming soon | Coming soon |
| DeepSWE | Coming soon | 43.0% | Coming soon |
| Terminal-Bench 2.1 | Coming soon | 78.9% | Coming soon |
| SWE-Atlas Codebase QnA | Coming soon | 32.2% | Coming soon |
| SWE Multilingual | Coming soon | 78.1% | Coming soon |
| SWE-bench Verified | Coming soon | 80.2% | Coming soon |
| SciCode | Coming soon | 48.3% | Coming soon |

### Reasoning

- Winner: Not comparable
- GPT-5.4 public-lane score: 64.2 (#23/28)
- Beam 1.0 501B-A23B FP8 public-lane score: 85.8 (Unranked · 3 rankable rows)

| Benchmark | GPT-5.4 | Beam 1.0 501B-A23B FP8 | Winner |
|-----------|-----------|-----------|--------|
| ARC-AGI-2 | 74.0% | Coming soon | Coming soon |
| ARC-AGI-3 | 0.2% | Coming soon | Coming soon |
| ARC-AGI-1 | 93.67% | Coming soon | Coming soon |
| CritPt | Coming soon | 17.4% | Coming soon |
| LongBench v2 | Coming soon | 67.6% | Coming soon |

### Multimodal & Grounded

- Winner: Not comparable
- GPT-5.4 public-lane score: 73.1 (#30/54)
- Beam 1.0 501B-A23B FP8 public-lane score: Coming soon

| Benchmark | GPT-5.4 | Beam 1.0 501B-A23B FP8 | Winner |
|-----------|-----------|-----------|--------|
| MMMU-Pro | 81.2% | Coming soon | Coming soon |
| OfficeQA Pro | 53.2% | Coming soon | Coming soon |
| MMMU-Pro w/ Python | 82.1% | Coming soon | Coming soon |
| CharXiv | 82.8% | Coming soon | Coming soon |
| ERQA | 65.4% | Coming soon | Coming soon |
| SimpleVQA | 61.1% | Coming soon | Coming soon |
| ScreenSpot Pro | 85.4% | Coming soon | Coming soon |
| ZeroBench | 41.0% | Coming soon | Coming soon |
| MedXpertQA (MM) | 77.1% | Coming soon | Coming soon |

### Knowledge

- Winner: Not comparable
- GPT-5.4 public-lane score: 69 (Supported · #20/184)
- Beam 1.0 501B-A23B FP8 public-lane score: Coming soon

| Benchmark | GPT-5.4 | Beam 1.0 501B-A23B FP8 | Winner |
|-----------|-----------|-----------|--------|
| GPQA | 92.8% | Coming soon | Coming soon |
| HLE | 52.1% | Coming soon | Coming soon |
| HLE w/o tools | 39.8% | 36% | GPT-5.4 |
| GPQA-D | 92.8% | 90.3% | GPT-5.4 |
| HealthBench Hard | 40.1% | Coming soon | Coming soon |
| MedXpertQA (Text) | 59.6% | Coming soon | Coming soon |
| HealthBench Professional | 48.1% | Coming soon | Coming soon |

### Instruction Following

- Winner: Directional only
- GPT-5.4 public-lane score: 90.7 (#23/148)
- Beam 1.0 501B-A23B FP8 public-lane score: 89.8 (#28/148)

| Benchmark | GPT-5.4 | Beam 1.0 501B-A23B FP8 | Winner |
|-----------|-----------|-----------|--------|
| IFBench | Coming soon | 80.7% | Coming soon |

### Mathematics

- Winner: Not comparable
- GPT-5.4 public-lane score: 82 (Unranked · 2 rankable rows)
- Beam 1.0 501B-A23B FP8 public-lane score: 87.9 (Unranked · 1 rankable row)

| Benchmark | GPT-5.4 | Beam 1.0 501B-A23B FP8 | Winner |
|-----------|-----------|-----------|--------|
| FrontierMath v2 (Tiers 1-3) | 78.596% | Coming soon | Coming soon |
| FrontierMath v2 (Tier 4) | 49.000% | Coming soon | Coming soon |
| AIME26 | Coming soon | 98.0% | Coming soon |

## FAQ

### Which is better overall, GPT-5.4 or Beam 1.0 501B-A23B FP8?

Beam 1.0 501B-A23B FP8 is ahead overall on BenchLM right now.

### Where is the biggest gap between GPT-5.4 and Beam 1.0 501B-A23B FP8?

The widest category gap is in reasoning, where the averages are 64.2 for GPT-5.4 and 85.8 for Beam 1.0 501B-A23B FP8.

### How many benchmarks does GPT-5.4 cover on BenchLM?

GPT-5.4 currently has 42 sourced benchmark scores on BenchLM.

### How many benchmarks does Beam 1.0 501B-A23B FP8 cover on BenchLM?

Beam 1.0 501B-A23B FP8 currently has 19 sourced benchmark scores on BenchLM.

## Related Comparisons

- [GPT-5.4 vs GPT-5.4 Pro](/compare/gpt-5-4-vs-gpt-5-4-pro)
- [Beam 1.0 501B-A23B FP8 vs GPT-5.4 Pro](/compare/gpt-5-4-pro-vs-reflection-beam-fp8)
- [GPT-5.4 vs GPT-5.4 mini](/compare/gpt-5-4-vs-gpt-5-4-mini)
- [Beam 1.0 501B-A23B FP8 vs GPT-5.4 mini](/compare/gpt-5-4-mini-vs-reflection-beam-fp8)

## Explore More

- [GPT-5.4 profile](/models/gpt-5-4)
- [Beam 1.0 501B-A23B FP8 profile](/models/reflection-beam-fp8)
- [Compare Pricing](/llm-pricing)
- [Alternative Finder](/tools/alternative-finder)
- [LLM Selector](/tools/llm-selector)
- [Overall Rankings](/best/overall)