Key Specifications
| Vendor | meta |
|---|
| Version | 3.1-405b |
|---|
| Release Date | 2024-07-23 |
|---|
| Context Window | 128000 tokens |
|---|
| Input Modalities | text |
|---|
| Output Modalities | text |
|---|
| License | Llama 3 Community License |
|---|
| Documentation | https://llama.meta.com/docs/ |
|---|
Benchmark Performance
| Benchmark | Score | Unit | Evaluated At | Notes | Source |
|---|
| MMLU | 88.6 | % | 2024-07-23 | 5-shot | view |
| HUMANEVAL | 89 | pass@1 | 2024-07-23 | — | view |
| GSM8K | 89.2 | % | 2024-07-23 | 0-shot CoT | view |
| MATH | 73.8 | % | 2024-07-23 | 0-shot CoT | view |
| BBH | 82.9 | % | 2024-07-23 | 3-shot CoT | view |
Pricing
| Tier | Price | Currency |
|---|
| Input | $5 / Mtok | USD |
| Output | $15 / Mtok | USD |
| Cache Read | $0 / Mtok | USD |
| Cache Write | $0 / Mtok | USD |
Source:
https://www.together.ai/pricing
· as of 2024-08-01
Compliance
- Data Residency: self-host
- SOC2: ✗
- HIPAA: ✗
- GDPR: ✗
- ISO 27001: ✗
Llama 3.1 405B
Ikhtisar Model
Meta Llama 3.1 405B 是当前最大规模的开源权重模型之一，拥有 4050 亿参数，128K 上下文窗口，在 MMLU、HumanEval、MATH 等基准上接近闭源旗舰水平，支持自托管部署。
Spesifikasi Inti
| Vendor | Versi | Tanggal Rilis | Jendela Konteks | Modalitas Input | Modalitas Output | Lisensi |
|---|
| Meta | 3.1-405b | 2024-07-23 | 128K | text | text | Llama 3 Community License |
Kinerja Benchmark
| Benchmark | Skor | Satuan | Catatan |
|---|
| MMLU (Massive Multitask Language Understanding) | 88.6 | % | 5-shot |
| HumanEval | 89.0 | pass@1 | — |
| GSM8K (Grade School Math 8K) | 89.2 | % | 0-shot CoT |
| MATH | 73.8 | % | 0-shot CoT |
| BBH (BIG-Bench Hard) | 82.9 | % | 3-shot CoT |
Harga
| Input | Output | Baca Cache | Tulis Cache |
|---|
| — | — | — | — |
per juta token
Kelebihan
- MMLU score 88.6, strong knowledge reasoning.
- HumanEval 89.0, excellent code generation.
- GSM8K 89.2, robust math reasoning.
Kekurangan
Kasus Penggunaan
Referensi