Inflection-2 is a powerful AI model that has been trained on 5,000 NVIDIA H100 GPUs with impressive computing capacity. It has outperformed Google’s PaLM 2 Large model in various AI benchmarks such as MMLU, TriviaQA, HellaSwag, and GSM8k. Despite being larger in size, Inflection-2 is more efficient than its predecessor, Inflection-1, thanks to a transition to H100 GPUs and optimized inference implementation. This achievement brings us closer to creating a personal AI for everyone. The company takes safety and trustworthiness seriously and supports global efforts to ensure responsible use of this technology. Inflection-2’s performance on different benchmarks, including math and coding tasks, is promising, and further enhancements are possible through fine-tuning. If you’re interested in contributing to the mission of Inflection AI, the team welcomes you to join them.
https://inflection.ai/inflection-2