Llama-2 13B Chat

Llama-2 13B Chat is a 13-billion parameter language model developed by Meta as part of the Llama 2 family of open-weight models. It is a fine-tuned variant optimized for dialogue and conversational use cases, trained using reinforcement learning from human feedback (RLHF). The model was released in July 2023 alongside other sizes in the Llama 2 lineup, including 7B and 70B variants, and is hosted on Hugging Face under the identifier meta-llama/Llama-2-13b-chat-hf.

The 13B parameter size positions this model between the lighter 7B and the larger 70B variants, offering a middle ground in terms of computational requirements and language understanding depth. It supports a context window of 4096 tokens and a maximum response size of 2500 tokens. The model is suited for tasks involving multi-turn conversation, instruction following, and general-purpose text generation where a fully open-weight model is preferred. Note that this model has been marked as deprecated on MindStudio.