AI startup Inflection has unveiled its new language model, Inflection-2, surpassing Google's PaLM 2 across multiple benchmarks. With an impressive computational prowess achieved by utilizing over 5,000 NVIDIA GPUs, Inflection-2 distinguishes itself in reasoning, factual knowledge, and stylistic versatility.

This advancement signals a competitive edge for Inflection, as it outperformed PaLM 2 in tasks from the Multi-task Middle-school Language Understanding (MMLU) to the Grade School Math (GSM8k) benchmarks. Inflection's commitment to top-tier AI safety, as a signatory to the White House's AI commitments, pairs with its drive to innovate.

The company's forward thrust includes enhancing its personal assistant app, Pi, with this model and ambitions to leverage a 22,000 GPU supercluster for larger models. Such developments pose significant competition to technology giants and underscore the accelerating pace in the AI landscape.