Key Specifications
| Vendor | meta |
|---|
| Version | 3.3-70b |
|---|
| Release Date | 2024-12-06 |
|---|
| Context Window | 128000 tokens |
|---|
| Input Modalities | text |
|---|
| Output Modalities | text |
|---|
| License | Llama 3.3 Community License |
|---|
| Documentation | https://llama.meta.com/docs/ |
|---|
Benchmark Performance
| Benchmark | Score | Unit | Evaluated At | Notes | Source |
|---|
| MMLU | 83.4 | % | 2024-12-06 | 5-shot | view |
| HUMANEVAL | 87 | pass@1 | 2024-12-06 | — | view |
| GSM8K | 86.9 | % | 2024-12-06 | 0-shot CoT | view |
| MATH | 73.8 | % | 2024-12-06 | 0-shot CoT | view |
| BBH | 83.9 | % | 2024-12-06 | 3-shot CoT | view |
| GPQA | 52.8 | % | 2024-12-06 | 0-shot | view |
| IFEVAL | 79.3 | % | 2024-12-06 | prompt_strict | view |
| ARC | 93.9 | % | 2024-12-06 | challenge | view |
| MUSR | 62.3 | % | 2024-12-06 | 0-shot | view |
| WINOGRANDE | 86.8 | % | 2024-12-06 | 0-shot | view |
Pricing
| Tier | Price | Currency |
|---|
| Input | $0.9 / Mtok | USD |
| Output | $0.9 / Mtok | USD |
| Cache Read | $0 / Mtok | USD |
| Cache Write | $0 / Mtok | USD |
Source:
https://ai.meta.com/blog/
· as of 2024-12-06
Compliance
- Data Residency: self-host
- SOC2: ✗
- HIPAA: ✗
- GDPR: ✗
- ISO 27001: ✗
Llama 3.3 70B
Visão geral do modelo
Meta Llama 3.3 70B 开源旗舰, 128K 上下文, 性能超越 Llama 3.1 405B, 接近 GPT-4o 水平。
Especificações principais
| Fornecedor | Versão | Data de lançamento | Janela de contexto | Modalidades de entrada | Modalidades de saída | Licença |
|---|
| Meta | 3.3-70b | 2024-12-06 | 128K | text | text | Llama 3.3 Community License |
Desempenho em benchmarks
| Benchmark | Pontuação | Unidade | Notas |
|---|
| MMLU (Massive Multitask Language Understanding) | 83.4 | % | 5-shot |
| HumanEval | 87.0 | pass@1 | — |
| GSM8K (Grade School Math 8K) | 86.9 | % | 0-shot CoT |
| MATH | 73.8 | % | 0-shot CoT |
| BBH (BIG-Bench Hard) | 83.9 | % | 3-shot CoT |
| GPQA | 52.8 | % | 0-shot |
| IFEval | 79.3 | % | prompt_strict |
| ARC | 93.9 | % | challenge |
| MUSR | 62.3 | % | 0-shot |
| WinoGrande | 86.8 | % | 0-shot |
Preços
| Entrada | Saída | Leitura de cache | Escrita de cache |
|---|
| — | — | — | — |
por milhão de tokens
Pontos fortes
- MMLU score 83.4, strong knowledge reasoning.
- HumanEval 87.0, excellent code generation.
- GSM8K 86.9, robust math reasoning.
Pontos fracos
Casos de uso
Referências