Galileo has launched “Galileo Luna,” a suite of evaluation foundation models (EFMs) specifically designed to evaluate the performance of LLMs like OpenAI's GPT-4 and Google's Gemini Pro.

These Luna EFM models, which are LLMs, have been fine-tuned to detect hallucinations, data leakages, context quality errors, and malicious prompts. Benchmark tests showed Luna EFMs outperforming existing evaluation models by up to 20% in accuracy.

The company claims Luna EFMs are faster, more cost-effective, and more accurate than current methods, including human evaluations and other LLMs like GPT-4. The company CEO stated that Luna EFMs can evaluate responses at a scale necessary for enterprises, being 97% cheaper, 11x faster, and 18% more accurate than OpenAI’s GPT-3.5.

Analyst QuickTake: The development of Luna has been an important step for Galileo, a leading GenAI evaluation company since early 2021. In February 2024 , the company also introduced retrieval augmented generation (RAG) and Agent Analytics to enhance the creation and dependability of AI applications, aiming for more accurate and transparent AI responses.