Jurassic-2 Models by AI21 Labs
About
Jurassic-2 (J2), developed by AI21 Labs and launched in March 2023, represents a cutting-edge family of large language models (LLMs) tailored for text completion and conversational scenarios. Initially, the J2 family featured three primary models—Large, Grande, and Jumbo—before simplifying to Ultra, Mid, and Light, based on size and capabilities. Known for their superior zero-shot instruction-following, reduced latency, and support for multiple languages, these models were trained on an extensive dataset comprising internet text, leading to over 60 billion parameters in their most advanced version. Despite their smaller parameter count, J2 models have demonstrated impressive performance on numerous benchmarks. AI21 Labs made these models available through the AI21 Studio platform, offering APIs for tasks such as paraphrasing and summarization. However, it's important to note that the Jurassic-2 models are no longer available as of November 14, 2024.
Current Variants
Use-when guidance is derived from seed capabilities, context, release, and replacement fields.
Use when the workload needs 8k context, 60B parameters, and structured outputs.
Use when the workload needs 8k context, 17B parameters, and structured outputs.
Use when the workload needs 8k context and 7B parameters.
Use when the workload needs 8k context and 60B parameters.
Use when the workload needs 8k context, 178B parameters, and structured outputs.
Use when the workload needs 8k context, 17B parameters, and structured outputs.
| Model | Use when | Released | Signals | Status |
|---|---|---|---|---|
| Jurassic-2 Ultra | Use when the workload needs 8k context, 60B parameters, and structured outputs. | 2023-03 | 8k context60B parametersstructured outputs | Current |
| Jurassic-2 Mid | Use when the workload needs 8k context, 17B parameters, and structured outputs. | 2023-03 | 8k context17B parametersstructured outputs | Current |
| Jurassic-2 Light | Use when the workload needs 8k context and 7B parameters. | 2023-03 | 8k context7B parameters | Current |
| Jurassic-2 Chat | Use when the workload needs 8k context and 60B parameters. | 2023-03 | 8k context60B parameters | Current |
| AI21 Labs J2-Ultra | Use when the workload needs 8k context, 178B parameters, and structured outputs. | 2021-08 | 8k context178B parametersstructured outputs | Current |
| AI21 Labs J2-Mid | Use when the workload needs 8k context, 17B parameters, and structured outputs. | 2021-08 | 8k context17B parametersstructured outputs | Current |
Release Timeline
2 release groupsSpecifications(6 models)
| Model | Released | Context | Parameters | Structured Outputs |
|---|---|---|---|---|
| Jurassic-2 Ultra | 2023-03 | 8k | 60B | Yes |
| Jurassic-2 Mid | 2023-03 | 8k | 17B | Yes |
| Jurassic-2 Light | 2023-03 | 8k | 7B | No |
| Jurassic-2 Chat | 2023-03 | 8k | 60B | No |
| AI21 Labs J2-Ultra | 2021-08 | 8k | 178B | Yes |
| AI21 Labs J2-Mid | 2021-08 | 8k | 17B | Yes |
Available From(2 providers)
Pricing
| Model | Provider | Input / 1M | Output / 1M | Type |
|---|---|---|---|---|
| Jurassic-2 Light | AI21 Studio | $3 | $3 | Serverless |
| AI21 Labs J2-Mid | AWS Bedrock | $6 | $6 | Serverless |
| AI21 Labs J2-Ultra | AWS Bedrock | $9 | $9 | Serverless |
| Jurassic-2 Mid | AWS Bedrock | $12.5 | $12.5 | Serverless |
| Jurassic-2 Mid | AI21 Studio | $12.5 | $12.5 | Serverless |
| Jurassic-2 Chat | AI21 Studio | $12.5 | $12.5 | Serverless |
| Jurassic-2 Ultra | AWS Bedrock | $18.8 | $18.8 | Serverless |
| Jurassic-2 Ultra | AI21 Studio | $18.8 | $18.8 | Serverless |
Frequently Asked Questions
- What is Jurassic-2 used for?
- Jurassic-2 is used for structured outputs and chatbot and role-playing use cases. The family description and listed model capabilities point to those workloads as the best fit.
- How does Jurassic-2 compare to Jamba 1.7?
- Jurassic-2 by AI21 Labs is strongest where you need structured outputs, while Jamba 1.7 by AI21 Labs is the closest related family to check for agent workflows and tool use. Jurassic-2 has 6 listed variants and reaches up to 8k context, while Jamba 1.7 reaches up to 256k context, so compare the specs and pricing tables before choosing a production model.
- Which Jurassic-2 model should I use?
- For the lowest listed input price, start with Jurassic-2 Light through AI21 Studio at $3/1M input tokens. For the most capable/latest local choice, evaluate AI21 Labs J2-Ultra with 8k context and structured outputs.
