litert-community

/

Qwen2.5-Coder-3B-Instruct

like 4

Follow

LiteRT Community (FKA TFLite) 8.64k

Thanks for converting Qwen2.5-Coder-3B — it answers correctly on both backends. This PR adds a measured Performance section to the card. Measured with the litert-lm CLI (benchmark -p 256 -d 256 --runs 3 --cache no) on an idle Apple M4 Max, generation-gated first (the backend produced correct text before any number was recorded). Happy to adjust the format if you'd like these to read differently.

Added on-device rows since opening this: a Galaxy S26 GPU-vs-CPU subsection (litert-lm 0.16.0, conditions inline, full delegation confirmed, both backends generation-checked), grouped with the M4 Max table under one Performance heading.