Llama 3.1 70B Instruct vs Qwen2.5 72B Instruct
Lado a lado. Cada métrica. Cada benchmark.
| Tipo | Llama 3.1 70B Instruct | Qwen2.5 72B Instruct |
|---|---|---|
| Provider |  Meta |  Alibaba Qwen |
| pontuação média | 37.8 | 53.2 |
| Preço de entrada | $0.40 | $0.12 |
| Preço de saída | $0.40 | $0.39 |
| Janela de contexto | 131K tokens (~66 books) | 33K tokens (~16 books) |
| Lançado em | 2024-07-23 | 2024-09-19 |
| Código aberto | Open Source | Open Source |
Pontuações de benchmark
15 benchmarks · Llama 3.1 70B Instruct: 4, Qwen2.5 72B Instruct: 11
| Benchmark | Categoria | Llama 3.1 70B Instruct | Qwen2.5 72B Instruct |
|---|---|---|---|
| Aider — Code Editing | coding | 58.6 | 65.4 |
| Balrog | knowledge | 27.9 | 16.2 |
| BBH (HuggingFace) | general | 55.9 | 61.9 |
| Chatbot Arena Elo — Overall | arena | 1292.8 | 1302.3 |
| CMMLU | knowledge | 64.4 | 85.7 |
| GPQA | knowledge | 14.2 | 16.7 |
| GPQA diamond | knowledge | 25.6 | 32.2 |
| IFEval | language | 86.7 | 86.4 |
| MATH level 5 | math | 36.7 | 63.2 |
| MATH Level 5 | math | 38.1 | 59.8 |
| MMLU | knowledge | 73.5 | 80.4 |
| MMLU-PRO | knowledge | 47.9 | 51.4 |
| MUSR | reasoning | 17.7 | 11.7 |
| OTIS Mock AIME 2024-2025 | math | 3.5 | 8.0 |
| The Agent Company | agentic | 6.9 | 5.7 |