Trase has reclaimed the top position on the Hugging Face GAIA leaderboard, achieving a published test score of nearly 67% and a validation score of 70.3%. This performance outpaces industry giants like Google, Microsoft, and OpenAI while costing roughly 1/100th per query.

  • Trase achieved a 66.78% overall test score on the GAIA benchmark, which evaluates multi-step reasoning, multimodal data handling, and tool usage.
  • The system spent approximately $10 in computing power per query, compared to OpenAI's Deep Research which cost between $1,100 and $1,300 per question.
  • Trase answers most questions in a single pass, whereas competitors like Deep Research required 64 attempts per question to aggregate their scores.
  • The platform utilizes fine-tuning on industry-specific data and outcome-supervised reinforcement learning to ensure accuracy and compliance in sectors like healthcare and finance.

Trase's model-agnostic architecture allows organizations to deploy secure, scalable AI agents without extensive IT teams or expensive infrastructure.