Key Specifications
| Vendor | other |
|---|
| Version | j2-ultra |
|---|
| Release Date | 2023-01-12 |
|---|
| Context Window | 8192 tokens |
|---|
| Input Modalities | text |
|---|
| Output Modalities | text |
|---|
| License | Proprietary |
|---|
| Documentation | https://huggingface.co/models |
|---|
Benchmark Performance
| Benchmark | Score | Unit | Evaluated At | Notes | Source |
|---|
| MMLU | 62.5 | % | 2023-01-12 | 5-shot | view |
| HUMANEVAL | 50.9 | pass@1 | 2023-01-12 | — | view |
| GSM8K | 30.3 | % | 2023-01-12 | 0-shot CoT | view |
| MATH | 27.5 | % | 2023-01-12 | 0-shot CoT | view |
| BBH | 52.9 | % | 2023-01-12 | 3-shot CoT | view |
| GPQA | 28.7 | % | 2023-01-12 | 0-shot | view |
| IFEVAL | 47.9 | % | 2023-01-12 | prompt_strict | view |
| ARC | 82.1 | % | 2023-01-12 | challenge | view |
| MUSR | 42.9 | % | 2023-01-12 | 0-shot | view |
| WINOGRANDE | 74.4 | % | 2023-01-12 | 0-shot | view |
Pricing
| Tier | Price | Currency |
|---|
| Input | $2 / Mtok | USD |
| Output | $2 / Mtok | USD |
| Cache Read | $0 / Mtok | USD |
| Cache Write | $0 / Mtok | USD |
Source:
https://huggingface.co/models
· as of 2023-01-12
Compliance
- Data Residency: self-host
- SOC2: ✗
- HIPAA: ✗
- GDPR: ✗
- ISO 27001: ✗
Jurassic-2 Ultra
Modellöversikt
AI21 Labs Jurassic-2 Ultra 旗舰模型, 8K 上下文, 在 J2 系列中性能最强, 适合复杂推理任务。
Kärnspecifikationer
| Leverantör | Version | Releasedatum | Kontextfönster | Inmatningsmodaliteter | Utmatningsmodaliteter | Licens |
|---|
| Other | j2-ultra | 2023-01-12 | 8K | text | text | Proprietary |
Benchmarkprestanda
| Benchmark | Poäng | Enhet | Anteckningar |
|---|
| MMLU (Massive Multitask Language Understanding) | 62.5 | % | 5-shot |
| HumanEval | 50.9 | pass@1 | — |
| GSM8K (Grade School Math 8K) | 30.3 | % | 0-shot CoT |
| MATH | 27.5 | % | 0-shot CoT |
| BBH (BIG-Bench Hard) | 52.9 | % | 3-shot CoT |
| GPQA | 28.7 | % | 0-shot |
| IFEval | 47.9 | % | prompt_strict |
| ARC | 82.1 | % | challenge |
| MUSR | 42.9 | % | 0-shot |
| WinoGrande | 74.4 | % | 0-shot |
Priser
| Inmatning | Utmatning | Cacheläsning | Cacheskrivning |
|---|
| — | — | — | — |
per miljon tokens
Styrkor
Svagheter
- 闭源专有模型，不支持自托管。
- 上下文窗口 8K 偏小。
Användningsområden
Referenser