Great Telugu model — mobile deployment perspective
#1
by 3morixd - opened
గొప్ప మోడల్!
We're looking for Telugu models that run on mobile. At Dispatch AI (UAE), we test on 40 phones (Snapdragon 865).
Challenge with Telugu: tokenizers often consume 2-3x more tokens for non-English text, meaning slower mobile generation.
We'd love to see token efficiency metrics and GGUF quantization results.
- Dispatch AI (FZE), Sharjah UAE | Building mobile AI for every language