WizardLM Models by WizardLM Team
About
The WizardLM family of large language models (LLMs) is based on the LLaMA architecture and utilizes a novel training method called Evol-Instruct. This method involves using AI to iteratively evolve and enhance instruction data, making it more complex and varied. The goal is to improve the model's ability to follow intricate instructions and handle complex tasks. By fine-tuning LLaMA with this AI-generated instruction data, WizardLM models have demonstrated superior performance compared to other LLaMA-based models trained on simpler data. Notably, WizardLM has shown competitive results in human evaluations and automatic assessments, sometimes even outperforming OpenAI's ChatGPT in handling high-complexity instructions.
Current Variants
Use-when guidance is derived from seed capabilities, context, release, and replacement fields.
Use when the workload needs 4k context and 70B parameters.
Use when the workload needs 2k context and 30B parameters.
Use when the workload needs 2k context, 13B parameters, and structured outputs.
Use when the workload needs 2k context and 7B parameters.
Use when the workload needs 2k context and 13B parameters.
Use when the workload needs 4k context and 13B parameters.
| Model | Use when | Released | Signals | Status |
|---|---|---|---|---|
| WizardLM 70B | Use when the workload needs 4k context and 70B parameters. | 2023-04 | 4k context70B parameters | Current |
| WizardLM 30B | Use when the workload needs 2k context and 30B parameters. | 2023-04 | 2k context30B parameters | Current |
| WizardLM 13B V1.0 | Use when the workload needs 2k context, 13B parameters, and structured outputs. | 2023-04 | 2k context13B parametersstructured outputs | Current |
| WizardLM 7B | Use when the workload needs 2k context and 7B parameters. | 2023-04 | 2k context7B parameters | Current |
| WizardLM 13B V1.1 | Use when the workload needs 2k context and 13B parameters. | 2023-04 | 2k context13B parameters | Current |
| WizardLM 13B V1.2 | Use when the workload needs 4k context and 13B parameters. | 2023-04 | 4k context13B parameters | Current |
Release Timeline
1 release groupSpecifications(6 models)
| Model | Released | Context | Parameters | Structured Outputs |
|---|---|---|---|---|
| WizardLM 70B | 2023-04 | 4k | 70B | No |
| WizardLM 30B | 2023-04 | 2k | 30B | No |
| WizardLM 13B V1.0 | 2023-04 | 2k | 13B | Yes |
| WizardLM 7B | 2023-04 | 2k | 7B | No |
| WizardLM 13B V1.1 | 2023-04 | 2k | 13B | No |
| WizardLM 13B V1.2 | 2023-04 | 4k | 13B | No |
Available From(3 providers)
Pricing
| Model | Provider | Input / 1M | Output / 1M | Type |
|---|---|---|---|---|
| WizardLM 13B V1.0 | Together AI | $0.3 | $0.3 | Serverless |
| WizardLM 13B V1.1 | Microsoft Foundry | $0.81 | $0.94 | Provisioned |
Frequently Asked Questions
- What is WizardLM used for?
- WizardLM is used for structured outputs, coding, and math-heavy prompts. The family description and listed model capabilities point to those workloads as the best fit.
- How does WizardLM compare to WizardCoder?
- WizardLM by WizardLM Team is strongest where you need structured outputs, while WizardCoder by WizardLM Team is the closest related family to check for coding. WizardLM has 6 listed variants and reaches up to 4k context, while WizardCoder reaches up to 100k context, so compare the specs and pricing tables before choosing a production model.
- Which WizardLM model should I use?
- For the lowest listed input price, start with WizardLM 13B V1.0 through Together AI at $0.3/1M input tokens. For the most capable/latest local choice, evaluate WizardLM 13B V1.0 with 2k context and structured outputs.
