Model Registry

Every LiteRT build the packages run, each pinned to a Hugging Face commit, and for every platform and backend what is known about it. The same data is in models.json. Install a build by name, with no URL to copy:

await FlutterEdgeAi.installModelFromCatalog(LiteRtLmModels.gemma4E2bIt.generic).install();

A blank cell is a combination that does not exist: a build compiled for one NPU on another platform, a web export on a phone, or a native build on the web. Package versions: flutter_edge_ai 2.5.4, flutter_edge_ai_litertlm 1.19.0, flutter_edge_ai_speech 0.5.12.

Language models#

DeepSeek R1 Distill Qwen 1.5B#

deepseek-r1-distill-qwen-1.5b · mit · Hugging Face

FastVLM 0.5B#

fastvlm-0.5b · apple-amlr · Hugging Face

• Documented: the Models page (at 0c626b2).

FunctionGemma 270M (Flutter demo fine-tune)#

functiongemma-270m-flutter-demo · unspecified · Hugging Face

FunctionGemma 270M#

functiongemma-270m-it · unspecified · Hugging Face

• Documented: the Models page (at 0c626b2).

Gemma 3 1B#

gemma-3-1b-it · gemma · Hugging Face

• Documented: the Models page (at 0c626b2).
• Documented: PR #697, run on a DDP Galaxy Tab S10+ (MT6989).
• Documented: PR #701, Gemma 3 NPU prefill chunks (LiteRT-LM#3508).

Gemma 3 270M#

gemma-3-270m-it · gemma · Hugging Face

• Documented: the Models page (at 0c626b2).

Gemma 3n E2B#

gemma-3n-e2b-it · gemma · Hugging Face

• Documented: the Models page (at 0c626b2).

Gemma 3n E4B#

gemma-3n-e4b-it · gemma · Hugging Face

• Documented: the Models page (at 0c626b2).

Gemma 4 E2B#

gemma-4-e2b-it · apache-2.0 · Hugging Face

• Documented: PR #688, device matrix (native-v0.18.0-f).
• Stale on macOS: MacBook Pro (M4 Pro), 2026-10-20, flutter_edge_ai_litertlm 1.16.1.
• Documented: PR #602, VENTUNO Q run (native-v0.18.0-c, litertlm 1.11.0).
• Documented: the Models page (at 0c626b2).

Gemma 4 E4B#

gemma-4-e4b-it · apache-2.0 · Hugging Face

• Documented: the Models page (at 0c626b2).
• Documented: the Models page (at 0c626b2).

LFM2.5 230M#

lfm2.5-230m · other · Hugging Face

• Documented: the Models page (at 0c626b2).

LLaVA-OneVision 0.5B#

llava-onevision-0.5b · apache-2.0 · Hugging Face

• Documented: the Models page (at 0c626b2).

Phi-4 Mini Instruct#

phi-4-mini-instruct · mit · Hugging Face

• Documented: the Models page (at 0c626b2).

Phi-4 Mini Reasoning#

phi-4-mini-reasoning · mit · Hugging Face

• Documented: the Models page (at 0c626b2).

Qwen2-VL 2B#

qwen2-vl-2b · apache-2.0 · Hugging Face

• Documented: the Models page (at 0c626b2).

Qwen 2.5 1.5B Instruct#

qwen2.5-1.5b-instruct · apache-2.0 · Hugging Face

• Documented: the Models page (at 0c626b2).

Qwen3 0.6B#

qwen3-0.6b · apache-2.0 · Hugging Face

• Documented: the Models page (at 0c626b2).
• Withheld: The catalog offers the dynamic INT4 build. This file failed to load from flutter_edge_ai_litertlm 1.7.0 until the repo replaced its tokenizer on 2026-10-13; the bytes pinned here are the replacement, which loads on macOS (litertlm 1.15.0) and has not been run elsewhere. A copy downloaded before 2026-10-13 still fails. Issue.

SmolLM3 3B#

smollm3-3b · apache-2.0 · Hugging Face

• Documented: the Models page (at 0c626b2).

SmolVLM2 500M#

smolvlm2-500m · apache-2.0 · Hugging Face

• Documented: the Models page (at 0c626b2).

TranslateGemma 4B#

translategemma-4b-it · other · Hugging Face

• Broken: On GPU the model emits only <pad> tokens. Issue.
• Documented: the Models page (at 0c626b2).

Embedding models#

EmbeddingGemma 2 Text + Vision + Audio 740M#

embeddinggemma-2-740m · apache-2.0 · Hugging Face

• Documented: the Models page (at 0c626b2).
• Documented: the Models page (at 0c626b2).
• Documented: the Embeddings & RAG page (at da9a3cb).

EmbeddingGemma 2 Text 270M#

embeddinggemma-2-text-270m · apache-2.0 · Hugging Face

• Documented: PR #688, device matrix (native-v0.18.0-f).
• Stale on macOS: MacBook Pro (M4 Pro), 2026-10-20, flutter_edge_ai_litertlm 1.16.1.
• Documented: the Models page (at 0c626b2).
• Documented: PR #702, Panther Lake run (native-v0.18.0-g).
• Documented: the Embeddings & RAG page (at da9a3cb).
• Documented: the Embeddings & RAG page (at da9a3cb).

EmbeddingGemma 2 Text + Vision 440M#

embeddinggemma-2-text-vision-440m · apache-2.0 · Hugging Face

• Documented: PR #688, device matrix (native-v0.18.0-f).
• Documented: the Models page (at 0c626b2).
• Documented: the Embeddings & RAG page (at da9a3cb).
• Documented: the Embeddings & RAG page (at da9a3cb).

EmbeddingGemma 300M#

embeddinggemma-300m · gemma · Hugging Face

• Documented: the Models page (at 0c626b2).

Gecko 110M (English)#

gecko-110m-en · apache-2.0 · Hugging Face

• Documented: the Models page (at 0c626b2).

Speech to text#

Moonshine Tiny#

moonshine-tiny · mit · Hugging Face

• Documented: the Models page (at 0c626b2).
• Verified on macOS: MacBook Pro (M4 Pro), 2026-10-20, flutter_edge_ai_speech 0.5.10.
• Documented: PR #688, device matrix (native-v0.18.0-f).

Parakeet CTC 0.6B#

parakeet-ctc-0.6b · cc-by-4.0 · Hugging Face

• Documented: PR #752, Galaxy S23 (Snapdragon 8 Gen 2), one 5 s window.
• Documented: the Models page (at 0c626b2).

Whisper Base#

whisper-base · apache-2.0 · Hugging Face

• Documented: the Models page (at 0c626b2).

Whisper Tiny#

whisper-tiny · apache-2.0 · Hugging Face

• Documented: the Models page (at 0c626b2).

Text to speech#

Inflect-Nano-v2#

inflect-nano-v2 · apache-2.0 · Hugging Face

• Documented: the Models page (at 0c626b2).

Matcha-TTS#

matcha-tts · mit · Hugging Face

• Documented: the Models page (at 0c626b2).
• Verified on macOS: MacBook Pro (M4 Pro), 2026-10-20, flutter_edge_ai_speech 0.5.10.
• Documented: PR #688, device matrix (native-v0.18.0-f).

Qwen3-TTS 0.6B#

qwen3-tts-0.6b · apache-2.0 · Hugging Face

• Documented: PR #695 (2026-10-19).