Model comparison
GPT-4 vs Llama 2-70B
GPT-4 is the stronger model overall, scoring 29.1 to 24.4 on the Noometry Index.
Last verified . 29 shared benchmarks.
MetaSummary
- They share 29 benchmarks with published results for both. GPT-4 scores higher in 8 categories and Llama 2-70B in 0 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in multilingual, where GPT-4 leads 40.6 to 27.7.
- The biggest single-benchmark swing is DTBench: 62.7% for GPT-4 and 41.6% for Llama 2-70B.
- Llama 2-70B has downloadable open weights; the other is API-only.
Side by side
| GPT-4 | Llama 2-70B | |
|---|---|---|
| Provider | OpenAI | Meta |
| Noometry Index | 29.1 | 24.4 |
| Released | 2023-03-14 | 2023-07-18 |
| Weights | Proprietary | Open |
| Context window | 8K | — |
| Max output | 8K | — |
| Input $ / M tokens | $30 | — |
| Output $ / M tokens | $60 | — |
| Results tracked | 38 | 35 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Too close to call
GPT-4: 31.6 (#283), Llama 2-70B: 31.4 (#286)
| Benchmark | GPT-4 | Llama 2-70B |
|---|---|---|
| LMArena Coding | 1254 | 1079 |
| WeirdML | 12.4% | — |
| BigCodeBench Instruct | 46% | — |
| BigCodeBench Complete | 57.2% | — |
| HumanEval+ | 79.3% | — |
Agentic & Tool Use Not comparable
GPT-4: —, Llama 2-70B: —
| Benchmark | GPT-4 | Llama 2-70B |
|---|---|---|
| METR Time Horizons | 36.1% | — |
Reasoning GPT-4 leads
GPT-4: 17.8 (#289), Llama 2-70B: 14.4 (#325)
| Benchmark | GPT-4 | Llama 2-70B |
|---|---|---|
| LMArena Hard Prompts | 1241 | 1073 |
| DTBench | 62.7% | 41.6% |
| BIG-Bench Hard | 75.1% | 64.9% |
| Epoch Capabilities Index | 125.89 | 113.79 |
| ForecastBench | 57.8 | 51.4 |
| HellaSwag | 95.3% | 85.3% |
| WinoGrande | 87.5% | 80.2% |
| Chess Puzzles | 4% | — |
| Mystery Game Puzzles | 12% | — |
| LMCA | 17.1% | — |
| CommonsenseQA 2.0 | — | 50% |
| LAMBADA | — | 78.9% |
| PIQA | — | 82.8% |
Math GPT-4 leads
GPT-4: 10.8 (#309), Llama 2-70B: 8.1 (#326)
| Benchmark | GPT-4 | Llama 2-70B |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 1.1% | 0% |
| LMArena Math | 1269 | 1091 |
| MATH Level 5 | 23% | 3.3% |
| GSM8K | 92% | 69.6% |
Knowledge GPT-4 leads
GPT-4: 18.4 (#282), Llama 2-70B: 7.4 (#310)
| Benchmark | GPT-4 | Llama 2-70B |
|---|---|---|
| GPQA Diamond | 35.7% | 26.3% |
| LMArena Expert | 1211 | 1039 |
| MMLU | 86.4% | 69.9% |
| TriviaQA | 84.8% | 87.6% |
| ARC (AI2) Challenge | — | 78.3% |
| BoolQ | — | 88.6% |
| OpenBookQA | — | 60.2% |
Multilingual GPT-4 leads
GPT-4: 40.6 (#215), Llama 2-70B: 27.7 (#274)
| Benchmark | GPT-4 | Llama 2-70B |
|---|---|---|
| LMArena Non-English | 1246 | 1045 |
| LMArena Chinese | 1242 | 995 |
| LMArena French | 1283 | 1090 |
| LMArena German | 1251 | 1041 |
| LMArena Japanese | 1209 | 927 |
| LMArena Korean | 1184 | 964 |
| LMArena Russian | 1251 | 1083 |
| LMArena Spanish | 1261 | 1143 |
Instruction Following GPT-4 leads
GPT-4: 65.3 (#222), Llama 2-70B: 54.9 (#278)
| Benchmark | GPT-4 | Llama 2-70B |
|---|---|---|
| LMArena Instruction Following | 1241 | 1071 |
Long Context GPT-4 leads
GPT-4: 37.7 (#212), Llama 2-70B: 32.3 (#270)
| Benchmark | GPT-4 | Llama 2-70B |
|---|---|---|
| LMArena Longer Query | 1244 | 1062 |
Writing & Preference GPT-4 leads
GPT-4: 34.9 (#268), Llama 2-70B: 32.3 (#279)
| Benchmark | GPT-4 | Llama 2-70B |
|---|---|---|
| LMArena Text | 1263 | 1115 |
| LMArena Creative Writing | 1244 | 1075 |
| LMArena Multi-Turn | 1257 | 1088 |
| EQ-Bench Creative Writing | 752 | — |
Frequently asked questions
Is GPT-4 better than Llama 2-70B?
GPT-4 is the stronger model overall, scoring 29.1 to 24.4 on the Noometry Index.
Is GPT-4 or Llama 2-70B better for coding?
They score almost the same on coding (31.6 vs 31.4); test both on your own repository before choosing.
How many benchmarks do GPT-4 and Llama 2-70B share?
29 benchmarks have published results for both models. GPT-4 has 38 scored results on Noometry and Llama 2-70B has 35.