Monthly Report

AI Hallucination Report

September 2026

191 benchmarked answers across 9 models this month.

191
Answers Benchmarked
9
Models Tested
3
High-Risk Models
1
Domains Covered

Worst-Performing Models

#1
Mistral Large
24 answers · avg correctness —
0
avg trust
#2
Qwen 2.5 72B
24 answers · avg correctness 88%
42
avg trust
#3
DeepSeek V3
29 answers · avg correctness 83%
47
avg trust
#4
GPT-5 Mini
9 answers · avg correctness 0%
50
avg trust
#5
Gemini 3 Flash
9 answers · avg correctness 0%
50
avg trust

Top Hallucinations This Month

trust 0Mistral Large

What is the current record for quantum entanglement distance achieved in a laboratory setting?

→ —

trust 0Qwen 2.5 72B

What is the largest prime number currently known?

→ —

trust 0DeepSeek V3

What is the largest prime number currently known?

→ —

trust 0Mistral Large

What is the largest prime number currently known?

→ —

trust 0Mistral Large

Does P equal NP?

→ —

Domain Breakdown

191
General