Skip to content
IFAI

Vectara

Company2 stories

  1. AI toolsReach 34

    Vectara's new Corrector cuts enterprise AI hallucinations to 0.9%

    Vectara just integrated a new Hallucination Corrector directly into its AI platform to catch and fix unreliable model responses in real time. Paired with its popular HHEM evaluator, the tool explains why a statement is inaccurate and rewrites it, reducing hallucination rates to about 0.9% in enterprise tests.

  2. IndustryReach 39

    The AI models that hallucinate the least

    Vectara's benchmark shows Zhipu AI's GLM-4-9B-Chat and Google's Gemini-2.0-Flash-Exp tied for the lowest hallucination rate at just 1.3%, followed by OpenAI's o1-mini at 1.4%. The numbers prove that smaller, specialized models can easily match or beat giant frontier LLMs when it comes to factual accuracy.