Great Turkish model — mobile deployment perspective
#51
by 3morixd - opened
Harika model!
We're always looking for Turkish models that can run on mobile. At Dispatch AI (UAE), we test models on 40 phones (Snapdragon 865).
Challenge with Turkish: tokenizers often consume 2-3x more tokens for non-English text, meaning slower generation on mobile.
We'd love to see: token efficiency metrics and GGUF quantization results for Turkish.
- Dispatch AI (FZE), Sharjah UAE | Building mobile AI for every language