# GPT-4 vs Llama 3.1-70B

> GPT-4 and Llama 3.1-70B score almost the same on the Noometry Index (28.9 vs 29.4), so choose on price, context window or the category you care about most.

- Canonical page: https://noometry.com/compare/gpt-4-vs-llama-3-1-70b
- Last updated: 2026-10-10
- Shared benchmarks: 28

## Summary

- They share 28 benchmarks with published results for both. GPT-4 scores higher in 4 categories and Llama 3.1-70B in 4 categories; 5 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where Llama 3.1-70B leads 24.1 to 18.4.
- The biggest single-benchmark swing is MATH Level 5: 23% for GPT-4 and 36.7% for Llama 3.1-70B.
- Llama 3.1-70B is cheaper at $0.40 / $0.40 per million input/output tokens, against $30 / $60 for GPT-4.
- Llama 3.1-70B accepts more context: 128K tokens versus 8K.
- Llama 3.1-70B has downloadable open weights; the other is API-only.

## Snapshot

| | GPT-4 | Llama 3.1-70B |
|---|---|---|
| Provider | OpenAI | Meta |
| Noometry Index | 28.9 | 29.4 |
| Rank | 308 | 299 |
| Context | 8K | 128K |
| Input $/M | $30 | $0.40 |
| Output $/M | $60 | $0.40 |
| Weights | Proprietary | Open |

## Coding

- GPT-4: 31.2 (#277)
- Llama 3.1-70B: 29.8 (#292)

| Benchmark | GPT-4 | Llama 3.1-70B |
|---|---|---|
| WeirdML | 12.4% | 9% |
| BigCodeBench Instruct | 46% | 46.1% |
| LMArena Coding | 1254 | 1260 |
| BigCodeBench Complete | 57.2% | 54.8% |
| HumanEval+ | 79.3% | — |

## Agentic & Tool Use

- GPT-4: —
- Llama 3.1-70B: 24.9 (#104)

| Benchmark | GPT-4 | Llama 3.1-70B |
|---|---|---|
| TheAgentCompany | — | 6.9% |
| BALROG | — | 27.9% |
| METR Time Horizons | 36.1% | — |

## Reasoning

- GPT-4: 17.3 (#285)
- Llama 3.1-70B: 20.8 (#215)

| Benchmark | GPT-4 | Llama 3.1-70B |
|---|---|---|
| LMArena Hard Prompts | 1241 | 1241 |
| DTBench | 62.7% | 60% |
| LMCA | 17.1% | 14.8% |
| Epoch Capabilities Index | 125.88 | 125.91 |
| Chess Puzzles | 4% | — |
| Mystery Game Puzzles | 12% | — |
| BIG-Bench Hard | 75.1% | — |
| ForecastBench | 57.8 | — |
| HellaSwag | 95.3% | — |
| WinoGrande | 87.5% | — |

## Math

- GPT-4: 10.1 (#304)
- Llama 3.1-70B: 12.7 (#298)

| Benchmark | GPT-4 | Llama 3.1-70B |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 1.1% | 3.6% |
| LMArena Math | 1269 | 1252 |
| MATH Level 5 | 23% | 36.7% |
| Omni-MATH | — | 21% |
| GSM8K | 92% | — |

## Knowledge

- GPT-4: 18.4 (#277)
- Llama 3.1-70B: 24.1 (#264)

| Benchmark | GPT-4 | Llama 3.1-70B |
|---|---|---|
| GPQA Diamond | 35.7% | 44.2% |
| LMArena Expert | 1211 | 1209 |
| MMLU | 86.4% | 80.1% |
| MMLU-Pro | — | 65.3% |
| GPQA (HELM) | — | 42.6% |
| TriviaQA | 84.8% | — |

## Multilingual

- GPT-4: 40.8 (#210)
- Llama 3.1-70B: 38.9 (#220)

| Benchmark | GPT-4 | Llama 3.1-70B |
|---|---|---|
| LMArena Non-English | 1246 | 1219 |
| LMArena Chinese | 1242 | 1215 |
| LMArena French | 1282 | 1260 |
| LMArena German | 1251 | 1222 |
| LMArena Japanese | 1209 | 1132 |
| LMArena Korean | 1184 | 1140 |
| LMArena Russian | 1251 | 1234 |
| LMArena Spanish | 1261 | 1253 |

## Instruction Following

- GPT-4: 65.5 (#217)
- Llama 3.1-70B: 65.5 (#218)

| Benchmark | GPT-4 | Llama 3.1-70B |
|---|---|---|
| LMArena Instruction Following | 1241 | 1231 |
| IFEval | — | 82.1% |

## Long Context

- GPT-4: 37.7 (#207)
- Llama 3.1-70B: 37.6 (#208)

| Benchmark | GPT-4 | Llama 3.1-70B |
|---|---|---|
| LMArena Longer Query | 1244 | 1241 |

## Writing & Preference

- GPT-4: 35.0 (#263)
- Llama 3.1-70B: 35.6 (#262)

| Benchmark | GPT-4 | Llama 3.1-70B |
|---|---|---|
| LMArena Text | 1263 | 1261 |
| LMArena Creative Writing | 1244 | 1232 |
| EQ-Bench Creative Writing | 752 | 784 |
| LMArena Multi-Turn | 1257 | 1256 |
| WildBench | — | 75.8% |

## FAQ

### Is GPT-4 better than Llama 3.1-70B?

GPT-4 and Llama 3.1-70B score almost the same on the Noometry Index (28.9 vs 29.4), so choose on price, context window or the category you care about most.

### Which is cheaper, GPT-4 or Llama 3.1-70B?

Llama 3.1-70B is cheaper. It lists at $0.40 per million input tokens and $0.40 per million output tokens; GPT-4 lists at $30 and $60.

### Is GPT-4 or Llama 3.1-70B better for coding?

GPT-4 scores higher on coding benchmarks: 31.2 versus 29.8 in the Noometry coding category.

### Which has the bigger context window?

Llama 3.1-70B does, with 128K tokens against 8K.

### How many benchmarks do GPT-4 and Llama 3.1-70B share?

28 benchmarks have published results for both models. GPT-4 has 38 scored results on Noometry and Llama 3.1-70B has 35.