I'm happy to share that today we are releasing the Mockingbird LLM. At <10B parameters it's an LLM trained to provide optimal results for RAG and structured ou

I'm happy to share that today we are releasing the Mockingbird LLM. At <10B parameters it's an LLM trained to provide optimal results for RAG and structured outputs. Although significantly smaller (and thus faster) than GPT-4 or Gemini-1.5 Pro, it performs at a comparable level for generation, citations, and structured outputs.