The xai-org/grok-3 model achieved a hallucination rate of 5.8%, a factual consistency rate of 94.2%, an answer rate of 93.0%, and an average summary length of 95.9 words as of March 20, 2026.

Authors

• Person: Not available Organization: Vectara
  vectara/hallucination-leaderboard - GitHub

Sources

• vectara/hallucination-leaderboard - GitHub github.com Vectara via serper

Referenced by nodes (4)

• hallucination rate concept
• answer rate concept
• factual consistency rate concept
• average summary length concept