Model comparison
Claude 3.5 Haiku vs GPT-4
Claude 3.5 Haiku and GPT-4 score almost the same on the Noometry Index (29.2 vs 29.1), so choose on price, context window or the category you care about most.
Last verified . 27 shared benchmarks.
Summary
- They share 27 benchmarks with published results for both. Claude 3.5 Haiku scores higher in 5 categories and GPT-4 in 3 categories; 4 gaps are clear of the uncertainty.
- The widest gap is in writing & preference, where Claude 3.5 Haiku leads 42.7 to 34.9.
- The biggest single-benchmark swing is MATH Level 5: 46.4% for Claude 3.5 Haiku and 23% for GPT-4.
Side by side
| Claude 3.5 Haiku | GPT-4 | |
|---|---|---|
| Provider | Anthropic | OpenAI |
| Noometry Index | 29.2 | 29.1 |
| Released | 2024-10-22 | 2023-03-14 |
| Weights | Proprietary | Proprietary |
| Context window | — | 8K |
| Max output | — | 8K |
| Input $ / M tokens | — | $30 |
| Output $ / M tokens | — | $60 |
| Results tracked | 49 | 38 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Claude 3.5 Haiku leads
Claude 3.5 Haiku: 32.9 (#265), GPT-4: 31.6 (#283)
| Benchmark | Claude 3.5 Haiku | GPT-4 |
|---|---|---|
| WeirdML | 30.7% | 12.4% |
| BigCodeBench Instruct | 46.1% | 46% |
| LMArena Coding | 1286 | 1254 |
| BigCodeBench Complete | 59% | 57.2% |
| Aider Polyglot | 28% | — |
| SciCode | 27.4% | — |
| LiveBench Coding | 51.4% | — |
| CadEval | 32% | — |
| HumanEval+ | — | 79.3% |
Agentic & Tool Use Not comparable
Claude 3.5 Haiku: 28.0 (#95), GPT-4: —
| Benchmark | Claude 3.5 Haiku | GPT-4 |
|---|---|---|
| BALROG | 19.3% | — |
| METR Time Horizons | — | 36.1% |
Reasoning Too close to call
Claude 3.5 Haiku: 17.7 (#290), GPT-4: 17.8 (#289)
| Benchmark | Claude 3.5 Haiku | GPT-4 |
|---|---|---|
| LMArena Hard Prompts | 1251 | 1241 |
| DTBench | 56.7% | 62.7% |
| Epoch Capabilities Index | 127.15 | 125.89 |
| CritPt | 0% | — |
| Chess Puzzles | — | 4% |
| LiveBench Reasoning | 28.1% | — |
| Mystery Game Puzzles | — | 12% |
| LiveBench Data Analysis | 48.5% | — |
| LMCA | — | 17.1% |
| BIG-Bench Hard | — | 75.1% |
| ForecastBench | — | 57.8 |
| HellaSwag | — | 95.3% |
| LiveBench | 43.5% | — |
| WinoGrande | — | 87.5% |
Math Claude 3.5 Haiku leads
Claude 3.5 Haiku: 14.7 (#300), GPT-4: 10.8 (#309)
| Benchmark | Claude 3.5 Haiku | GPT-4 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 4.3% | 1.1% |
| LMArena Math | 1244 | 1269 |
| MATH Level 5 | 46.4% | 23% |
| Omni-MATH | 22.4% | — |
| LiveBench Math | 35.5% | — |
| FrontierMath (Feb 2025 set) | 0.3% | — |
| GSM8K | — | 92% |
Knowledge Too close to call
Claude 3.5 Haiku: 18.7 (#281), GPT-4: 18.4 (#282)
| Benchmark | Claude 3.5 Haiku | GPT-4 |
|---|---|---|
| GPQA Diamond | 38.1% | 35.7% |
| LMArena Expert | 1208 | 1211 |
| MMLU | 74.3% | 86.4% |
| MMLU-Pro | 60.5% | — |
| Confabulations | 36.7% | — |
| GPQA (HELM) | 36.3% | — |
| TriviaQA | — | 84.8% |
Multimodal Not comparable
Claude 3.5 Haiku: 26.8 (#117), GPT-4: —
| Benchmark | Claude 3.5 Haiku | GPT-4 |
|---|---|---|
| LMArena Vision | 1092 | — |
| GeoBench | 34% | — |
Multilingual Too close to call
Claude 3.5 Haiku: 40.0 (#218), GPT-4: 40.6 (#215)
| Benchmark | Claude 3.5 Haiku | GPT-4 |
|---|---|---|
| LMArena Non-English | 1238 | 1246 |
| LMArena Chinese | 1229 | 1242 |
| LMArena French | 1264 | 1283 |
| LMArena German | 1237 | 1251 |
| LMArena Japanese | 1175 | 1209 |
| LMArena Korean | 1173 | 1184 |
| LMArena Russian | 1253 | 1251 |
| LMArena Spanish | 1261 | 1261 |
Instruction Following GPT-4 leads
Claude 3.5 Haiku: 62.9 (#234), GPT-4: 65.3 (#222)
| Benchmark | Claude 3.5 Haiku | GPT-4 |
|---|---|---|
| LMArena Instruction Following | 1241 | 1241 |
| LiveBench Instruction Following | 61.9% | — |
| IFEval | 79.2% | — |
Long Context Too close to call
Claude 3.5 Haiku: 38.3 (#200), GPT-4: 37.7 (#212)
| Benchmark | Claude 3.5 Haiku | GPT-4 |
|---|---|---|
| LMArena Longer Query | 1261 | 1244 |
Writing & Preference Claude 3.5 Haiku leads
Claude 3.5 Haiku: 42.7 (#234), GPT-4: 34.9 (#268)
| Benchmark | Claude 3.5 Haiku | GPT-4 |
|---|---|---|
| LMArena Text | 1255 | 1263 |
| LMArena Creative Writing | 1233 | 1244 |
| EQ-Bench Creative Writing | 1146 | 752 |
| LMArena Multi-Turn | 1265 | 1257 |
| Short-Story Creative Writing | 73.5% | — |
| WildBench | 76% | — |
| LiveBench Language | 35.4% | — |
Frequently asked questions
Is Claude 3.5 Haiku better than GPT-4?
Claude 3.5 Haiku and GPT-4 score almost the same on the Noometry Index (29.2 vs 29.1), so choose on price, context window or the category you care about most.
Is Claude 3.5 Haiku or GPT-4 better for coding?
Claude 3.5 Haiku scores higher on coding benchmarks: 32.9 versus 31.6 in the Noometry coding category.
How many benchmarks do Claude 3.5 Haiku and GPT-4 share?
27 benchmarks have published results for both models. Claude 3.5 Haiku has 49 scored results on Noometry and GPT-4 has 38.