A side-by-side comparison of release history, capabilities, licensing and model sizes.
| Attribute |
 Vicuna
|
 WizardLM
|
|---|---|---|
| Developer | UC Berkeley, CMU, Stanford, MBZUAI, UCSD | WizardLM |
| Current generation | – | – |
| Initial release | 2023-03-30 | 2023-05-26 |
| Instruct tuned | Yes | Yes |
| Coding | No | No |
| Finetuning | Yes | Yes |
| Open source | Yes | Yes |
| License | Noncommercial | Noncommercial |
| Model sizes | 13B | 7B, 13B, 70B, 8x22B |
| Links |
Released alongside Koala, Vicuna is one of many descendants of the Meta LLaMA model trained on dialogue data collected from the ShareGPT website. According to the authors, Vicuna achieves more than 90% of ChatGPT's quality in user preference tests, while vastly outperforming Alpaca. As of May 2023, Vicuna seems to be the heir apparent of the instruct-finetuned LLaMA model family, though it is also restricted from commercial use.
WizardLM is a variant of LLaMA trained with complex instructions.
WizardLM is a LLM based on LLaMA trained using a new method, called Evol-Instruct, on complex instruction data. By using AI to "evolve" instructions, WizardLM outperforms similar LLaMA-based LLMs trained on simpler instruction data. On the 6th of July, 2023, WizardLM V1.1 was released with significantly improved performance, and as of 15 April 2024, WizardLM-2 was released with state-of-the-art performance.
