Fact · Voice and AI
24.2percent
On the same Vectara leaderboard table, the highest hallucination rate is 24.2 percent, for the mistralai/ministral-3-3b-2512 model, which answered only 74.3 percent of prompts.
Checked against its source on October 17, 2026
In the source's own words
“mistralai/ministral-3-3b-2512|24.2 %|75.8 %|74.3 %|119.4”
Vectara Hallucination Leaderboard: LLM hallucination rates when summarizing a document, Vectara, Vectara, GitHub, 2026, Leaderboard table, lower rows.
Open the original- What it measures
- The table runs from the lowest rate, 1.8 percent, to this highest rate of 24.2 percent, with the models ranked by hallucination rate.
- What it does not show
- A small model that refused many prompts; the answer rate is shown in the same row. Rates describe one dataset and one prompt in the September 2026 snapshot, so they are not a ranking of the models for other tasks.