From the model card
Fields below are copied from the tags and counters on the HuggingFace repository litert-community/gemma-4-E4B-it-litert-lm at our last fetch. They are set by the uploader, not verified by us; rows with no tag are omitted. How this page is made.
- Publisher (HF namespace)
- litert-community
- License tag
apache-2.0— read the license file in the repo before relying on it- Lineage
-
- base model google/gemma-4-E4B-it
- fine-tune of google/gemma-4-E4B-it
- Downloads (HF counter at last fetch)
- 375,851
- Likes (HF counter at last fetch)
- 190
- Model card
- https://huggingface.co/litert-community/gemma-4-E4B-it-litert-lm
Use cases
- Running Gemma 4 instruction following natively on Android devices
- Building on-device AI features in mobile apps via Google's LiteRT-LM API
- Privacy-preserving mobile AI where data cannot leave the device
- Offline AI assistant functionality in mobile applications
- Research on on-device LLM deployment constraints and quality trade-offs
Pros
- Official Google LiteRT-LM format for seamless Android integration
- E4B selective activation reduces per-step compute for mobile constraints
- 98 likes; reflects developer interest in on-device Gemma 4 deployment
- Avoids cloud API costs for mobile AI features in consumer apps
Cons
- Requires LiteRT-LM runtime; not compatible with standard inference frameworks
- E4B on-device quality is lower than full Gemma 4 server inference
- Limited to high-end mobile hardware; low-end devices will not meet memory requirements
- LiteRT-LM API is Android-first; iOS support is less mature
Tags
litert-lmbase_model:google/gemma-4-E4B-itbase_model:finetune:google/gemma-4-E4B-itlicense:apache-2.0region:us