← All models

DeepSeek LLM 7B Chat

DeepSeek · 2023-11-29 · 6.9B parameters

DeepSeek LLM 7B Chat is the instruction-tuned checkpoint of DeepSeek LLM 7B Base, released November 29, 2023; DeepSeek reports supervised fine-tuning on about 1.5 million English and Chinese instances followed by DPO. It keeps the base model's 30-layer multi-head-attention architecture and 4,096-token context, with 6,910,365,696 parameters computed from its configuration.

Benchmark scores