Google's new Gemini Pro is worse than GPT-3

Gemini is the latest addition to Google DeepMind's series of large language models (LLMs). It stands out as it marks the first instance where reported results rival the performance of the OpenAI GPT model series across a diverse range of tasks. Specifically, Gemini's "Ultra" version is said to outshine GPT-4 on various tasks, while the "Pro" version is deemed comparable to GPT-3.5. However, the lack of detailed evaluation information and model predictions being made public hinders the ability to replicate, …

artificial deepmind diverse gemini gemini pro google google deepmind gpt gpt-3 gpt-4 instance language language models large language large language models llms marks openai openai gpt performance series tasks