BREAKING
Mistral under 40% on propaganda test
0
models tested
0
questions
0
narratives
How the test worked
1
75 questions
↓
2
3 languages
↓
3
Score 1 to 5
↓
4
No web search
0
top Mistral rank
0
cited RT or Sputnik
0
versions under 40%
Top performers vs Mistral
Top group
Leaders
●
Claude (Anthropic)
●
Some Chinese models
●
Grok
Lower group
Mistral
●
All four versions under 40%
●
Cited sanctioned sources
●
Europe's AI champion
Limits of standalone LLMs
AI NEWS BLITZ
An Estonian study finds Mistral models struggle to spot Russian propaganda.