Part 3/13:
According to Olivier Goon, lead of OpenAI's API platform, this move aligns with the company's mission to "make AI more accessible for everyone." The model supports both text and vision inputs, with future plans to incorporate multimodal outputs like video and audio. Based on benchmark scores, GPT-40 Mini demonstrates impressive capabilities, scoring 82% on the Massive Multitask Language Understanding (MMLU) benchmark—significantly higher than GPT-3.5’s 70%, and approaching GPT-4's 88.7%. While Google's Gemini Ultra leads with a 90% score, GPT-40 Mini's performance is promising for a smaller model.