Just months after it announced its first AI module, artificial intelligence company Inflection AI has announced Inflection 2, which it calls, “The best model in the world for its compute class and the second most capable LLM in the world today.”
In benchmark tests, Inflection-2 showcases its prowess by outperforming Google’s latest model on scientific question-answer datasets and surpassing Meta’s model on multiple benchmarks. The competitive edge in common sense and mathematical reasoning sets Inflection-2 apart.
“Our new model, Inflection-2, is substantially more capable than Inflection-1, demonstrating much improved factual knowledge, better stylistic control, and dramatically improved reasoning,” Inflection AI wrote.
Despite not focusing on coding and mathematical reasoning during training, Inflection-2 surprises with its promising abilities in these fields. The model’s versatility also proves crucial for addressing diverse AI applications.
Integrating Inflection-2 into Pi
The company is gearing up to integrate Inflection-2 into its personal AI assistant chatbot, Pi. The move promises not only enhanced efficiency but also signifies the practical application of the language model.
Also, Inflection AI plans to scale its models using an extensive GPU cluster, placing a key emphasis on responsible scaling. The company also recognises the ethical considerations in AI development, setting a precedent for mindful advancements.
“Designed with serving efficiency in mind, Inflection-2 will soon be powering Pi. Thanks to a transition from A100 to H100 GPUs, as well as our highly optimized inference implementation, we managed to reduce the cost and increase the speed of serving vs. Inflection-1 despite Inflection-2 being multiple times larger,” the company wrote.
In a landscape dominated by industry giants like OpenAI and Nvidia, startups like Inflection will have to strategically optimise efficiency to impact AI models even on a smaller scale. However, the company’s unveiling of Inflection-2 signals a new wave of competition in the AI space.
