DEEPSEEK-V2 (HUGGING FACE HUB)

Summary from the TLDR AI newsletter

DeepSeek has released a 200B+ parameter model with 21B active parameters. It performs extremely well on code and reasoning. It's not clear if it is overall better than Llama 3 70B, but it is a welcome addition to the open model ecosystem.