Model comparison
Llama 2-70B vs Llama 3-70B
Llama 3-70B is the stronger model overall, scoring 28.5 to 24.1 on the Noometry Index.
Last verified . 25 shared benchmarks.
MetaSummary
- They share 25 benchmarks with published results for both. Llama 2-70B scores higher in 0 categories and Llama 3-70B in 8 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where Llama 3-70B leads 20.8 to 7.4.
- The biggest single-benchmark swing is MATH Level 5: 3.3% for Llama 2-70B and 22.6% for Llama 3-70B.
Side by side
| Llama 2-70B | Llama 3-70B | |
|---|---|---|
| Provider | Meta | Meta |
| Noometry Index | 24.1 | 28.5 |
| Released | 2023-07-18 | 2024-04-18 |
| Weights | Open | Open |
| Context window | — | — |
| Max output | — | — |
| Input $ / M tokens | — | — |
| Output $ / M tokens | — | — |
| Results tracked | 35 | 31 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Llama 3-70B leads
Llama 2-70B: 30.8 (#281), Llama 3-70B: 35.3 (#215)
| Benchmark | Llama 2-70B | Llama 3-70B |
|---|---|---|
| LMArena Coding | 1079 | 1206 |
| BigCodeBench Instruct | — | 43.6% |
| BigCodeBench Complete | — | 54.5% |
| HumanEval+ | — | 72% |
| MBPP+ | — | 69% |
Agentic & Tool Use Not comparable
Llama 2-70B: —, Llama 3-70B: 20.9 (#132)
| Benchmark | Llama 2-70B | Llama 3-70B |
|---|---|---|
| Cybench | — | 5% |
Reasoning Llama 3-70B leads
Llama 2-70B: 13.7 (#324), Llama 3-70B: 17.4 (#284)
| Benchmark | Llama 2-70B | Llama 3-70B |
|---|---|---|
| LMArena Hard Prompts | 1073 | 1195 |
| DTBench | 41.6% | 54.2% |
| Epoch Capabilities Index | 113.59 | 122.91 |
| ForecastBench | 51.4 | 57.1 |
| WinoGrande | 80.2% | 83.5% |
| Kagi LLM Benchmark | — | 35.1% |
| BIG-Bench Hard | 64.9% | — |
| CommonsenseQA 2.0 | 50% | — |
| HellaSwag | 85.3% | — |
| LAMBADA | 78.9% | — |
| PIQA | 82.8% | — |
Math Llama 3-70B leads
Llama 2-70B: 7.5 (#320), Llama 3-70B: 12.0 (#299)
| Benchmark | Llama 2-70B | Llama 3-70B |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 0% | 4.3% |
| LMArena Math | 1091 | 1218 |
| MATH Level 5 | 3.3% | 22.6% |
| GSM8K | 69.6% | — |
Knowledge Llama 3-70B leads
Llama 2-70B: 7.4 (#305), Llama 3-70B: 20.8 (#272)
| Benchmark | Llama 2-70B | Llama 3-70B |
|---|---|---|
| GPQA Diamond | 26.3% | 40.6% |
| LMArena Expert | 1039 | 1149 |
| MMLU | 69.9% | 79.3% |
| ARC (AI2) Challenge | 78.3% | — |
| BoolQ | 88.6% | — |
| OpenBookQA | 60.2% | — |
| TriviaQA | 87.6% | — |
Multilingual Llama 3-70B leads
Llama 2-70B: 27.9 (#269), Llama 3-70B: 33.8 (#246)
| Benchmark | Llama 2-70B | Llama 3-70B |
|---|---|---|
| LMArena Non-English | 1045 | 1142 |
| LMArena Chinese | 994 | 1114 |
| LMArena French | 1089 | 1231 |
| LMArena German | 1041 | 1169 |
| LMArena Japanese | 927 | 1017 |
| LMArena Korean | 964 | 1017 |
| LMArena Russian | 1083 | 1159 |
| LMArena Spanish | 1142 | 1241 |
Instruction Following Llama 3-70B leads
Llama 2-70B: 55.1 (#273), Llama 3-70B: 62.8 (#233)
| Benchmark | Llama 2-70B | Llama 3-70B |
|---|---|---|
| LMArena Instruction Following | 1071 | 1194 |
Long Context Llama 3-70B leads
Llama 2-70B: 32.2 (#265), Llama 3-70B: 35.6 (#235)
| Benchmark | Llama 2-70B | Llama 3-70B |
|---|---|---|
| LMArena Longer Query | 1062 | 1174 |
Writing & Preference Llama 3-70B leads
Llama 2-70B: 32.5 (#274), Llama 3-70B: 43.0 (#225)
| Benchmark | Llama 2-70B | Llama 3-70B |
|---|---|---|
| LMArena Text | 1115 | 1221 |
| LMArena Creative Writing | 1075 | 1210 |
| LMArena Multi-Turn | 1088 | 1223 |
Frequently asked questions
Is Llama 2-70B better than Llama 3-70B?
Llama 3-70B is the stronger model overall, scoring 28.5 to 24.1 on the Noometry Index.
Is Llama 2-70B or Llama 3-70B better for coding?
Llama 3-70B scores higher on coding benchmarks: 35.3 versus 30.8 in the Noometry coding category.
How many benchmarks do Llama 2-70B and Llama 3-70B share?
25 benchmarks have published results for both models. Llama 2-70B has 35 scored results on Noometry and Llama 3-70B has 31.