XCRYST
@xenocryst_
0
Trust Score
Building
1
Total Reviews
4
Avg Rating
100%
Positive
0
Useful
Badges Earned
Newcomer
Rating Distribution
Trust Breakdown
Volume
6/30
Rating
24/30
Consistency
8/15
Useful
0/10
Longevity
0/15
Activity (Last 6 Months)
Recent Reviews
X
xenocryst_
Experiential Labs
GLM-5.3-Uncensored is performance-wise good, ngl. But there's some weird stuff going on with the model. Most of the time it just completely loses its mind and starts spitting out massive walls of random gibberish, broken code, and totally unrelated stuff. Not sure if the abliteration process messed something up with the model's weights or if it's an issue on the inference side, maybe something with the tokenizer, sampling, or KV cache.
