gemma4-e2b-rknn3-rk1828 by Firefly AI: Open-Weight Model

Model Card

By Firefly AI, published under apache-2.0, revision a0f25cce537c.

llamapi run handles model download, loading, and execution. See the LlamaPi Wiki for setup and usage. Gemma 4 E2B IT is a compact instruction-tuned multimodal model developed by Google DeepMind. - Accepts text, image, and audio inputs and generates text output with an encoder-free unified architecture. - Supports reasoning, coding, multilingual, and agentic tasks with a 256K context window and 140+ languages. - Designed as an efficient open model for on-device and edge deployment. The models provided in this repository are converted and adapted by the Firefly AI team from the original models. All intellectual property rights and copyright in the underlying original models remain with the…

Read Firefly AI's full model card

Gemma 4 E2B IT

RK1828 | LlamaPi Ready

Model Overview

Item	Details
Original model	Gemma 4 E2B IT (ModelScope \| Hugging Face)
Original developer	Google DeepMind
Model type	Multimodal generation
Target platform	RK1828
Converted by	Firefly AI Team
Deployment tool	LlamaPi

Run with LlamaPi

llamapi run gemma4:e2b

llamapi run handles model download, loading, and execution. See the LlamaPi Wiki for setup and usage.

About the Original Model

Gemma 4 E2B IT is a compact instruction-tuned multimodal model developed by Google DeepMind.

Accepts text, image, and audio inputs and generates text output with an encoder-free unified architecture.
Supports reasoning, coding, multilingual, and agentic tasks with a 256K context window and 140+ languages.
Designed as an efficient open model for on-device and edge deployment.

License & Attribution

The models provided in this repository are converted and adapted by the Firefly AI team from the original models. All intellectual property rights and copyright in the underlying original models remain with the original official team, and their distribution is governed by the corresponding official open-source license.

Configuration

Architecture: Gemma4ForConditionalGeneration
Context length (tokens): 131,072
Layers: 35
Hidden size: 1,536
Feed-forward size: 6,144
Attention heads: 8
Key/value heads: 1
Head dimension: 256
Vocabulary size: 262,144
Sliding window (tokens): 512
Model type: gemma4

Identity and Version

Repository: t-firefly/gemma4-e2b-rknn3-rk1828
Publisher: Firefly AI
Task: Any to any
Modality: Multimodal
Library: Not stated by the source
Parameters: Not stated by the source
Languages: Not stated by the source
Revision: a0f25cce537c7bf810fa153d8c4c2b7c64ec592d
First published: 2026-09-21
Last updated: 2026-09-30

Files and Weights

18 files, 7.9 GB in total. The weights are 4 files totalling 5.6 GB in bin, gguf, safetensors.

Weights4 files · 5.6 GB

Configuration1 file · 5.0 KB

Tokenizer2 files · 32.2 MB

Documentation1 file · 2.0 KB

Other9 files · 2.2 GB

Repository1 file · 2.0 KB

Every file

File	Type	Size	SHA-256
gemma-4-e2b-it.embed.bin	Weights	805.3 MB	c7a23c49fa7c
gemma-4-e2b-it.safetensors	Weights	125.8 MB	3927357464a1
gemma-4-e2b-it.tokenizer.gguf	Weights	15.8 MB	a3cc2a0296b9
gemma-4-e2b-it_per_layer_inputs.embed.bin	Weights	4.7 GB	f2a3b25ebdf9
config.json	Configuration	5.0 KB	—
README.md	Documentation	2.0 KB	—
audio_gemma-4-e2b-it.rknn	Other	10.7 MB	87d675534088
audio_gemma-4-e2b-it.weight	Other	626.8 MB	cf37d0057d9b
chat_template.jinja	Other	18.6 KB	—
firefly-logo-outline.png	Other	59.5 KB	—
llm_gemma-4-e2b-it-40960.rknn	Other	25.1 MB	1b0c5967a321
llm_gemma-4-e2b-it.weight	Other	1.4 GB	dfcb463500d5
model.toml	Other	200 B	—
vision_gemma-4-e2b-it-384.rknn	Other	5.1 MB	e3ea6e2239f5
vision_gemma-4-e2b-it-384.weight	Other	91.4 MB	820d87a8b133
.gitattributes	Repository	2.0 KB	—
tokenizer.json	Tokenizer	32.2 MB	cc8d3a0ce364
tokenizer_config.json	Tokenizer	3.1 KB	—

License and Download

License: apache-2.0
Access: Open weights, no gate
Download size: 5.6 GB

Download from Firefly AI

Released by Firefly AI through its official repository on Hugging Face. Read the license.

Built From

Derived from google/gemma-4-E2B-it
Quantized from google/gemma-4-E2B-it

Memory Requirements

Precision	Weights in memory
As published	5.6 GB

Weights only, from the published parameter count; the key-value cache and runtime add to this.

Questions About gemma4-e2b-rknn3-rk1828

Can I use gemma4-e2b-rknn3-rk1828 commercially?

Yes. gemma4-e2b-rknn3-rk1828 is released under Apache License 2.0. The Apache License 2.0 is a permissive open-source license. It permits commercial use, modification and redistribution. It requires keeping the license and copyright notices and any NOTICE file, stating significant changes, and it includes an express patent grant from contributors.

What is gemma4-e2b-rknn3-rk1828's context length?

131,072 tokens, from the maximum position embeddings in its published configuration.

Similar Models

Model · Any to any

gemma-4-12B-it-qat-GGUF

Unsloth AI

This model ships a Multi-Token Prediction drafter at the repo root (mtp-gemma-4-12B-it.gguf, a near-lossless smart Q40). A recent llama.cpp auto-discovers it from -hf, so you do not pass --model-draft: The drafter shares the target's KV cache and does not change the output (the target verifies every drafted token). See the MTP/ folder for the other precisions and explicit usage. Gemma is a family of open models built by Google DeepMind. Gemma 4 models are multimodal, handling text and image input (with audio supported on E2B, E4B, and 12B) and generating text output. This release includes open-weights models in both pre-trained and instruction-tuned variants. Gemma 4 features a context…

Open weights apache-2.0 transformers

View model

Model · Any to any

gemma-4-E4B-it-GGUF

GGML Org

Run with https://llama.app - https://huggingface.co/google/gemma-4-E4B-it - https://huggingface.co/google/gemma-4-E4B-it-assistant - https://huggingface.co/google/gemma-4-E4B-it-qat-q40-unquantized-assistant - https://huggingface.co/google/gemma-4-E4B-it-qat-q40-unquantized - add info - add dflash

Open weights apache-2.0

View model

Model · Any to any

gemma-4-12B-it-qat-q4_0-gguf

Google

Gemma is a family of open models built by Google DeepMind. Gemma 4 models are multimodal, handling text and image input (with audio supported on E2B, E4B, and 12B) and generating text output. This release includes open-weights models in both pre-trained and instruction-tuned variants. Gemma 4 features a context window of up to 256K tokens and maintains multilingual support in over 140 languages. Featuring both Dense and Mixture-of-Experts (MoE) architectures, Gemma 4 is well-suited for tasks like text generation, coding, and reasoning. The models are available in five distinct sizes: E2B, E4B, 12B, 26B A4B, and 31B. Their diverse sizes make them deployable in environments ranging from…

Open weights apache-2.0 transformers

View model

Model · Any to any

gemma-4-E4B-it-qat-q4_0-gguf

Google

Gemma is a family of open models built by Google DeepMind. Gemma 4 models are multimodal, handling text and image input (with audio supported on E2B, E4B, and 12B) and generating text output. This release includes open-weights models in both pre-trained and instruction-tuned variants. Gemma 4 features a context window of up to 256K tokens and maintains multilingual support in over 140 languages. Featuring both Dense and Mixture-of-Experts (MoE) architectures, Gemma 4 is well-suited for tasks like text generation, coding, and reasoning. The models are available in five distinct sizes: E2B, E4B, 12B, 26B A4B, and 31B. Their diverse sizes make them deployable in environments ranging from…

Open weights apache-2.0

View model

Model · Any to any

gemma-4-E4B-it-qat-GGUF

Unsloth AI

This model ships a Multi-Token Prediction drafter at the repo root (mtp-gemma-4-E4B-it.gguf, a near-lossless smart Q40). A recent llama.cpp auto-discovers it from -hf, so you do not pass --model-draft: The drafter shares the target's KV cache and does not change the output (the target verifies every drafted token). See the MTP/ folder for the other precisions and explicit usage. Gemma is a family of open models built by Google DeepMind. Gemma 4 models are multimodal, handling text and image input (with audio supported on E2B, E4B, and 12B) and generating text output. This release includes open-weights models in both pre-trained and instruction-tuned variants. Gemma 4 features a context…

Open weights apache-2.0 131,072 tokens transformers

View model

Model · Any to any

gemma-4-E2B-it-qat-GGUF

Unsloth AI

This model ships a Multi-Token Prediction drafter at the repo root (mtp-gemma-4-E2B-it.gguf, a near-lossless smart Q40). A recent llama.cpp auto-discovers it from -hf, so you do not pass --model-draft: The drafter shares the target's KV cache and does not change the output (the target verifies every drafted token). See the MTP/ folder for the other precisions and explicit usage. Gemma is a family of open models built by Google DeepMind. Gemma 4 models are multimodal, handling text and image input (with audio supported on E2B, E4B, and 12B) and generating text output. This release includes open-weights models in both pre-trained and instruction-tuned variants. Gemma 4 features a context…

Open weights apache-2.0 131,072 tokens transformers

View model