DeepSeek - AI Assistant
authentic ratingArguably the most capable model in the set, undermined by a recurring language-switching bug and a hard regeneration limit.
55real score
4store star
10,904ratings
About the rating
Store rating 3.99 is lower than textAvg 3.34 atypical for rating inflation. Detailed genuine reviews in Russian and English cover specific cases (school assignments, storytelling, emotional support). Criticism is equally specific: mid-conversation language switching, regeneration cap.
Strengths
What users love
- Users appreciate the transparent 'thinking' mode that shows its reasoning
- The depth of answers
- And the fact that the model pushes back instead of just agreeing. Russian-language reviews describe real emotional conversations.
Weak spots
What users hate
- The model frequently switches to Chinese mid-conversation and does not remember language preferences. A limit of 6 regenerations per conversation makes iterative work impossible. No line break support in the input field.
Who it is for
Researchers, students, and anyone who wants an AI that thinks rather than flatters.
Screenshots
