# Claude 3 Haiku vs GPT-5.4

> GPT-5.4 is the stronger model overall, scoring 59.5 to 25.7 on the Noometry Index.

- Canonical page: https://noometry.com/compare/claude-3-haiku-vs-gpt-5-4
- Last updated: 2026-10-10
- Shared benchmarks: 27

## Summary

- They share 27 benchmarks with published results for both. Claude 3 Haiku scores higher in 0 categories and GPT-5.4 in 9 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in math, where GPT-5.4 leads 73.3 to 9.2.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 1.8% for Claude 3 Haiku and 97.8% for GPT-5.4.

## Snapshot

| | Claude 3 Haiku | GPT-5.4 |
|---|---|---|
| Provider | Anthropic | OpenAI |
| Noometry Index | 25.7 | 59.5 |
| Rank | 333 | 14 |
| Context | — | 1.05M |
| Input $/M | — | $2.50 |
| Output $/M | — | $15 |
| Weights | Proprietary | Proprietary |

## Coding

- Claude 3 Haiku: 26.1 (#320)
- GPT-5.4: 52.8 (#28)

| Benchmark | Claude 3 Haiku | GPT-5.4 |
|---|---|---|
| WeirdML | 9.8% | 77.7% |
| LMArena Coding | 1199 | 1497 |
| SWE-bench Verified | — | 76.9% |
| DeepSWE | — | 51.8% |
| LMArena WebDev | — | 1465 |
| SciCode | — | 56.6% |
| GSO | — | 31.4% |
| BigCodeBench Instruct | 39.4% | — |
| MirrorCode | — | 15.6% |
| BigCodeBench Complete | 50.1% | — |
| CadEval | 12% | — |
| ALE-Bench | — | 1,607 |
| AlgoTune | — | 1.85 |
| HumanEval+ | 68.9% | — |
| MBPP+ | 68.8% | — |

## Agentic & Tool Use

- Claude 3 Haiku: —
- GPT-5.4: 46.5 (#12)

| Benchmark | Claude 3 Haiku | GPT-5.4 |
|---|---|---|
| Terminal-Bench | — | 81.8% |
| APEX-Agents | — | 52.4% |
| τ²-bench Banking | — | 39.4% |
| DeepResearch Bench | — | 35.1% |
| PostTrainBench | — | 19% |
| GBAEval | — | 45.1% |
| LMArena Search | — | 1197 |
| METR Time Horizons | — | 74.3% |
| Vending-Bench 2 | — | 6,144 |

## Reasoning

- Claude 3 Haiku: 15.7 (#303)
- GPT-5.4: 61.7 (#17)

| Benchmark | Claude 3 Haiku | GPT-5.4 |
|---|---|---|
| Kagi LLM Benchmark | 34.2% | 63.8% |
| LMArena Hard Prompts | 1174 | 1485 |
| DTBench | 50.1% | 94.4% |
| LMCA | 8.8% | 52% |
| Epoch Capabilities Index | 118.3 | 156.92 |
| ForecastBench | 53.2 | 59.5 |
| ARC-AGI-2 | — | 74% |
| NYT Connections (extended) | — | 91.3% |
| ARC-AGI-1 | — | 93.7% |
| CritPt | — | 23.4% |
| Chess Puzzles | — | 44% |
| EnigmaEval | — | 16% |
| Thematic Generalization | — | 80% |
| EBR-Bench | — | 25.4% |
| Mystery Game Puzzles | — | 37% |
| WinoGrande | 74.2% | — |

## Math

- Claude 3 Haiku: 9.2 (#313)
- GPT-5.4: 73.3 (#14)

| Benchmark | Claude 3 Haiku | GPT-5.4 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 1.8% | 97.8% |
| LMArena Math | 1188 | 1488 |
| FrontierMath (Tiers 1-3) | — | 78.6% |
| FrontierMath Tier 4 | — | 49% |
| MathArena Final-Answer Competitions | — | 83.1% |
| ProofBench | — | 56% |
| MATH Level 5 | 14.9% | — |
| FrontierMath (Feb 2025 set) | — | 47.6% |
| FrontierMath Tier 4 (v1) | — | 27.1% |

## Knowledge

- Claude 3 Haiku: 17.3 (#280)
- GPT-5.4: 65.3 (#13)

| Benchmark | Claude 3 Haiku | GPT-5.4 |
|---|---|---|
| GPQA Diamond | 36.3% | 93.3% |
| LMArena Expert | 1148 | 1507 |
| Humanity's Last Exam | — | 36.2% |
| SimpleQA Verified | — | 45.1% |
| Confabulations | 34.2% | — |
| Vectara Hallucination Rate | — | 7% |
| MMLU | 73.8% | — |

## Multimodal

- Claude 3 Haiku: 23.1 (#123)
- GPT-5.4: 43.8 (#19)

| Benchmark | Claude 3 Haiku | GPT-5.4 |
|---|---|---|
| LMArena Vision | 950 | 1302 |
| Blueprint-Bench 2 | — | 27.1% |
| Furniture Assembly | — | 37.5% |
| LMArena Document | — | 1471 |
| ScienceQA | 72% | — |

## Multilingual

- Claude 3 Haiku: 36.1 (#238)
- GPT-5.4: 56.4 (#22)

| Benchmark | Claude 3 Haiku | GPT-5.4 |
|---|---|---|
| LMArena Non-English | 1178 | 1465 |
| LMArena Chinese | 1155 | 1520 |
| LMArena French | 1194 | 1493 |
| LMArena German | 1174 | 1472 |
| LMArena Japanese | 1102 | 1484 |
| LMArena Korean | 1108 | 1449 |
| LMArena Russian | 1204 | 1480 |
| LMArena Spanish | 1165 | 1453 |

## Instruction Following

- Claude 3 Haiku: 61.5 (#242)
- GPT-5.4: 77.3 (#25)

| Benchmark | Claude 3 Haiku | GPT-5.4 |
|---|---|---|
| LMArena Instruction Following | 1174 | 1469 |

## Long Context

- Claude 3 Haiku: 36.1 (#232)
- GPT-5.4: 50.3 (#8)

| Benchmark | Claude 3 Haiku | GPT-5.4 |
|---|---|---|
| LMArena Longer Query | 1191 | 1474 |
| CL-bench | — | 27.9% |
| CL-bench Life | — | 21.7% |

## Writing & Preference

- Claude 3 Haiku: 29.8 (#286)
- GPT-5.4: 72.1 (#18)

| Benchmark | Claude 3 Haiku | GPT-5.4 |
|---|---|---|
| LMArena Text | 1195 | 1469 |
| LMArena Creative Writing | 1157 | 1439 |
| EQ-Bench Creative Writing | 717 | 1840 |
| LMArena Multi-Turn | 1190 | 1482 |
| EQ-Bench 4 | — | 1272 |

## FAQ

### Is Claude 3 Haiku better than GPT-5.4?

GPT-5.4 is the stronger model overall, scoring 59.5 to 25.7 on the Noometry Index.

### Is Claude 3 Haiku or GPT-5.4 better for coding?

GPT-5.4 scores higher on coding benchmarks: 52.8 versus 26.1 in the Noometry coding category.

### How many benchmarks do Claude 3 Haiku and GPT-5.4 share?

27 benchmarks have published results for both models. Claude 3 Haiku has 37 scored results on Noometry and GPT-5.4 has 68.