LLM Reference

WizardLM 70B

About

WizardLM 70B is an advanced large language model utilizing the transformer architecture with an impressive 70 billion parameters. It is adept at following complex instructions and managing multi-turn conversations with detailed and polite responses. Known for its capabilities in text and code generation, as well as solving mathematical problems, the model often surpasses some proprietary LLMs on specific benchmarks. Offered in various quantized formats like GGML and GGUF, WizardLM 70B is optimized for diverse hardware and performance requirements. Although the training data isn't publicly accessible, its open-source nature, coupled with availability on platforms such as Hugging Face, makes it a valuable resource for research and development. Specialized versions like WizardCoder and WizardMath focus on coding and mathematical tasks, respectively.

Capabilities

MultimodalFunction CallingTool UseJSON Mode

Specifications

FamilyWizardLM
Parameters70B
ArchitectureDecoder Only
Specializationgeneral