# Good and Small models for Mobile Devices 

Try them out in 

![PrivacyAIIcon](https://www.acmeup.com/icons/32.png) [Privacy AI](https://apps.apple.com/app/apple-store/id6738392421?pt=127450194&ct=huggingface&mt=8) 
on the App Store. 

[Privacy AI](https://apps.apple.com/app/apple-store/id6738392421?pt=127450194&ct=huggingface&mt=8) is a lightweight, serverless application. All tools - including web search, stock quotes, and Health analysis - run on-device, keeping data and actions fully private. It supports both local AI models and connections to your own OpenAI-compatible servers. 

Refer more information on [Privacy AI Official Site:](https://privacyai.acmeup.com)

## Qwen3 4B Instruct 2507
Qwen3-4B-Instruct-2507 is the latest 4B parameter model in the Qwen3 series, featuring significant improvements in reasoning, mathematics, science, coding, and tool usage. With 262K context length and strong multilingual support, it excels at instruction following, logical reasoning, and complex problem-solving tasks.  

**Model Intention:** Latest Qwen3-4B Instruct model with enhanced reasoning, logical thinking, mathematics, science, coding, and tool usage capabilities  

**Model URL:** [https://huggingface.co/flyingfishinwater/good_and_small_models/resolve/main/Qwen3-4B-Instruct-2507-Q4_0.gguf](https://huggingface.co/flyingfishinwater/good_and_small_models/resolve/main/Qwen3-4B-Instruct-2507-Q4_0.gguf)  

**Model Info URL:** [https://huggingface.co/Qwen/Qwen3-4B-Instruct-2507](https://huggingface.co/Qwen/Qwen3-4B-Instruct-2507)  

**Model License:** [License Info](https://www.apache.org/licenses/LICENSE-2.0.txt)  

**Model Description:** Qwen3-4B-Instruct-2507 is the latest 4B parameter model in the Qwen3 series, featuring significant improvements in reasoning, mathematics, science, coding, and tool usage. With 262K context length and strong multilingual support, it excels at instruction following, logical reasoning, and complex problem-solving tasks.  

**Developer:** [https://huggingface.co/Qwen](https://huggingface.co/Qwen)  

**File Size:** 2400 MB  

**Context Length:** 2048 tokens  

**Prompt Format:** 
```

```

**Template Name:** qwen  

**Add BOS Token:** Yes  

**Add EOS Token:** No  

**Parse Special Tokens:** Yes  


---

## Qwen3 4B Thinking 2507
Qwen3-4B-Thinking-2507 is a specialized variant of the Qwen3-4B series with enhanced reasoning capabilities. It features thinking mode enabled by default, providing significantly improved performance on complex reasoning tasks including logical reasoning, mathematics, science, coding, and academic benchmarks with 262K context length.  

**Model Intention:** Advanced reasoning model with thinking mode enabled for complex logical reasoning, mathematics, science, and coding tasks  

**Model URL:** [https://huggingface.co/flyingfishinwater/good_and_small_models/resolve/main/Qwen3-4B-Thinking-2507-Q4_0.gguf](https://huggingface.co/flyingfishinwater/good_and_small_models/resolve/main/Qwen3-4B-Thinking-2507-Q4_0.gguf)  

**Model Info URL:** [https://huggingface.co/Qwen/Qwen3-4B-Thinking-2507](https://huggingface.co/Qwen/Qwen3-4B-Thinking-2507)  

**Model License:** [License Info](https://www.apache.org/licenses/LICENSE-2.0.txt)  

**Model Description:** Qwen3-4B-Thinking-2507 is a specialized variant of the Qwen3-4B series with enhanced reasoning capabilities. It features thinking mode enabled by default, providing significantly improved performance on complex reasoning tasks including logical reasoning, mathematics, science, coding, and academic benchmarks with 262K context length.  

**Developer:** [https://huggingface.co/Qwen](https://huggingface.co/Qwen)  

**File Size:** 2100 MB  

**Context Length:** 2048 tokens  

**Prompt Format:** 
```

```

**Template Name:** qwen  

**Add BOS Token:** Yes  

**Add EOS Token:** No  

**Parse Special Tokens:** Yes  


---

## GLM Edge 4B Chat
GLM-4 is the latest generation of pre-trained models in the GLM-4 series launched by Zhipu AI. In the evaluation of data sets in semantics, mathematics, reasoning, code, and knowledge, GLM-4 has shown superior performance beyond Llama-3. In addition to multi-round conversations, GLM-4-Chat also has advanced features such as web browsing, code execution, custom tool calls (Function Call), and long text reasoning (supporting up to 128K context). This generation of models has added multi-language support, supporting 26 languages including Japanese, Korean, and German.  

**Model Intention:** It is the latest generation of pre-trained models in the GLM-4 series launched by Zhipu AI  

**Model URL:** [https://huggingface.co/flyingfishinwater/good_and_small_models/resolve/main/glm-edge-4b-chat.Q4_K_M.gguf](https://huggingface.co/flyingfishinwater/good_and_small_models/resolve/main/glm-edge-4b-chat.Q4_K_M.gguf)  

**Model Info URL:** [https://huggingface.co/THUDM](https://huggingface.co/THUDM)  

**Model License:** [License Info](https://huggingface.co/THUDM/glm-edge-4b-chat/raw/main/LICENSE)  

**Model Description:** GLM-4 is the latest generation of pre-trained models in the GLM-4 series launched by Zhipu AI. In the evaluation of data sets in semantics, mathematics, reasoning, code, and knowledge, GLM-4 has shown superior performance beyond Llama-3. In addition to multi-round conversations, GLM-4-Chat also has advanced features such as web browsing, code execution, custom tool calls (Function Call), and long text reasoning (supporting up to 128K context). This generation of models has added multi-language support, supporting 26 languages including Japanese, Korean, and German.  

**Developer:** [https://huggingface.co/THUDM](https://huggingface.co/THUDM)  

**File Size:** 2627 MB  

**Context Length:** 1024 tokens  

**Prompt Format:** 
```
{% for item in messages %}{% if item['role'] == 'system' %}<|system|>
{{ item['content'] }}{% elif item['role'] == 'user' %}<|user|>
{{ item['content'] }}{% elif item['role'] == 'assistant' %}<|assistant|>
{{ item['content'] }}{% endif %}{% endfor %}{% if add_generation_prompt %}<|assistant|>
{% endif %}
```

**Template Name:** glm  

**Add BOS Token:** Yes  

**Add EOS Token:** No  

**Parse Special Tokens:** Yes  


---

## Gemma 3n E2B it
Gemma 3n models are designed for efficient execution on low-resource devices. They are capable of multimodal input, handling text, image, video, and audio input, and generating text outputs, with open weights for pre-trained and instruction-tuned variants. These models were trained with data in over 140 spoken languages.  

**Model Intention:** Gemma 3n models are designed for efficient execution on low-resource devices. They are capable of multimodal input  

**Model URL:** [https://huggingface.co/flyingfishinwater/good_and_small_models/resolve/main/gemma-3n-E2B-it-Q4_0.gguf](https://huggingface.co/flyingfishinwater/good_and_small_models/resolve/main/gemma-3n-E2B-it-Q4_0.gguf)  

**Model Info URL:** [https://huggingface.co/google/gemma-3n-E2B-it](https://huggingface.co/google/gemma-3n-E2B-it)  

**Model License:** [License Info](https://choosealicense.com/licenses/apache-2.0/)  

**Model Description:** Gemma 3n models are designed for efficient execution on low-resource devices. They are capable of multimodal input, handling text, image, video, and audio input, and generating text outputs, with open weights for pre-trained and instruction-tuned variants. These models were trained with data in over 140 spoken languages.  

**Developer:** [https://huggingface.co/google](https://huggingface.co/google)  

**Update Date:** 2025-06-27  

**File Size:** 2720 MB  

**Context Length:** 4096 tokens  

**Prompt Format:** 
```

```

**Template Name:** chatml  

**Add BOS Token:** Yes  

**Add EOS Token:** No  

**Parse Special Tokens:** Yes  


---

## SmolLM3 3B
SmolLM3 is a fully open model that offers strong performance at the 3B–4B scale. The model is a decoder-only transformer using GQA and NoPE (with 3:1 ratio), it was pretrained on 11.2T tokens with a staged curriculum of web, code, math and reasoning data. Post-training included midtraining on 140B reasoning tokens.  

**Model Intention:** SmolLM3 is a 3B parameter language model designed to push the boundaries of small models. It supports 6 languages (English, French, Spanish, German, Italian, and Portuguese), advanced reasoning and long context.  

**Model URL:** [https://huggingface.co/flyingfishinwater/good_and_small_models/resolve/main/SmolLM3-Q4_K_M.gguf](https://huggingface.co/flyingfishinwater/good_and_small_models/resolve/main/SmolLM3-Q4_K_M.gguf)  

**Model Info URL:** [https://huggingface.co/HuggingFaceTB/SmolLM3-3B](https://huggingface.co/HuggingFaceTB/SmolLM3-3B)  

**Model License:** [License Info](https://www.apache.org/licenses/LICENSE-2.0.txt)  

**Model Description:** SmolLM3 is a fully open model that offers strong performance at the 3B–4B scale. The model is a decoder-only transformer using GQA and NoPE (with 3:1 ratio), it was pretrained on 11.2T tokens with a staged curriculum of web, code, math and reasoning data. Post-training included midtraining on 140B reasoning tokens.  

**Developer:** [https://huggingface.co/HuggingFaceTB](https://huggingface.co/HuggingFaceTB)  

**File Size:** 1920 MB  

**Context Length:** 2048 tokens  

**Prompt Format:** 
```

```

**Template Name:** qwen  

**Add BOS Token:** Yes  

**Add EOS Token:** No  

**Parse Special Tokens:** Yes  


---

## Phi4 mini 4B
Phi-4-mini-instruct is a lightweight open model built upon synthetic data and filtered publicly available websites - with a focus on high-quality, reasoning dense data. The model is intended for broad multilingual commercial and research use. The model provides uses for general purpose AI systems and applications which require: 1). Memory/compute constrained environments; 2). Latency bound scenarios; 3) Strong reasoning (especially math and logic).  The model is designed to accelerate research on language and multimodal models, for use as a building block for generative AI powered features.  

**Model Intention:** Phi-4-mini-instruct is a lightweight model focused on high-quality, reasoning dense data. It supports 128K token context length  

**Model URL:** [https://huggingface.co/flyingfishinwater/good_and_small_models/resolve/main/Phi-4-mini-instruct-Q4_K_M.gguf](https://huggingface.co/flyingfishinwater/good_and_small_models/resolve/main/Phi-4-mini-instruct-Q4_K_M.gguf)  

**Model Info URL:** [https://huggingface.co/microsoft/Phi-4-mini-instruct](https://huggingface.co/microsoft/Phi-4-mini-instruct)  

**Model License:** [License Info](https://choosealicense.com/licenses/mit/)  

**Model Description:** Phi-4-mini-instruct is a lightweight open model built upon synthetic data and filtered publicly available websites - with a focus on high-quality, reasoning dense data. The model is intended for broad multilingual commercial and research use. The model provides uses for general purpose AI systems and applications which require: 1). Memory/compute constrained environments; 2). Latency bound scenarios; 3) Strong reasoning (especially math and logic).  The model is designed to accelerate research on language and multimodal models, for use as a building block for generative AI powered features.  

**Developer:** [https://huggingface.co/microsoft](https://huggingface.co/microsoft)  

**File Size:** 2020 MB  

**Context Length:** 2048 tokens  

**Prompt Format:** 
```
{% for message in messages %}{% if message['role'] == 'system' and 'tools' in message and message['tools'] is not none %}{{ '<|' + message['role'] + '|>' + message['content'] + '<|tool|>' + message['tools'] + '<|/tool|>' + '<|end|>' }}{% else %}{{ '<|' + message['role'] + '|>' + message['content'] + '<|end|>' }}{% endif %}{% endfor %}{% if add_generation_prompt %}{{ '<|assistant|>' }}{% else %}{{ eos_token }}{% endif %}
```

**Template Name:** llama3.2  

**Add BOS Token:** Yes  

**Add EOS Token:** No  

**Parse Special Tokens:** Yes  


---

## Qwen3 1.7B
Qwen3 1.7B is one of the small models in the Qwen series, designed for efficiency and speed. It can run seamlessly on edge devices, enabling rapid inference and real-time applications. This compact model is ideal for testing scenarios, prototyping, or deployment in resource-constrained environments.  

**Model Intention:** The 1.7B model in the Qwen3 series is a small model designed for fast predictions and function calls.  

**Model URL:** [https://huggingface.co/flyingfishinwater/good_and_small_models/resolve/main/Qwen3-1.7B-Q4_K_M.gguf](https://huggingface.co/flyingfishinwater/good_and_small_models/resolve/main/Qwen3-1.7B-Q4_K_M.gguf)  

**Model Info URL:** [https://huggingface.co/Qwen/Qwen3-1.7B](https://huggingface.co/Qwen/Qwen3-1.7B)  

**Model License:** [License Info](https://www.apache.org/licenses/LICENSE-2.0.txt)  

**Model Description:** Qwen3 1.7B is one of the small models in the Qwen series, designed for efficiency and speed. It can run seamlessly on edge devices, enabling rapid inference and real-time applications. This compact model is ideal for testing scenarios, prototyping, or deployment in resource-constrained environments.  

**Developer:** [https://huggingface.co/Qwen](https://huggingface.co/Qwen)  

**File Size:** 1110 MB  

**Context Length:** 2048 tokens  

**Prompt Format:** 
```

```

**Template Name:** qwen  

**Add BOS Token:** Yes  

**Add EOS Token:** No  

**Parse Special Tokens:** Yes  


---

## ERNIE-4.5 0.3B
ERNIE 4.5 is a series of open source models created by Baidu. The advanced capabilities of the ERNIE 4.5 models, particularly the MoE-based A47B and A3B series, are underpinned by several key technical innovations: 1. Multimodal Heterogeneous MoE Pre-Training; 2. Scaling-Efficient Infrastructure; 3. Modality-Specific Post-Training  

**Model Intention:** ERNIE-4.5-0.3B-Base is a text dense Base model for testing the model's architecture.  

**Model URL:** [https://huggingface.co/flyingfishinwater/good_and_small_models/resolve/main/ERNIE-4.5-0.3B-PT-Q4_0.gguf](https://huggingface.co/flyingfishinwater/good_and_small_models/resolve/main/ERNIE-4.5-0.3B-PT-Q4_0.gguf)  

**Model Info URL:** [https://huggingface.co/baidu/ERNIE-4.5-0.3B-Base-PT](https://huggingface.co/baidu/ERNIE-4.5-0.3B-Base-PT)  

**Model License:** [License Info](https://www.apache.org/licenses/LICENSE-2.0.txt)  

**Model Description:** ERNIE 4.5 is a series of open source models created by Baidu. The advanced capabilities of the ERNIE 4.5 models, particularly the MoE-based A47B and A3B series, are underpinned by several key technical innovations: 1. Multimodal Heterogeneous MoE Pre-Training; 2. Scaling-Efficient Infrastructure; 3. Modality-Specific Post-Training  

**Developer:** [https://huggingface.co/baidu](https://huggingface.co/baidu)  

**File Size:** 233 MB  

**Context Length:** 2048 tokens  

**Prompt Format:** 
```

```

**Template Name:** qwen  

**Add BOS Token:** Yes  

**Add EOS Token:** No  

**Parse Special Tokens:** Yes  


---

## LFM2 1.2B
LFM2 is a new generation of hybrid models developed by Liquid AI, specifically designed for edge AI and on-device deployment. It sets a new standard in terms of quality, speed, and memory efficiency. LFM2 is a new hybrid Liquid model with multiplicative gates and short convolutions. It supported languages: English, Arabic, Chinese, French, German, Japanese, Korean, and Spanish.  

**Model Intention:** LFM2 1.2B is particularly suited for agentic tasks, data extraction, RAG, creative writing, and multi-turn conversations  

**Model URL:** [https://huggingface.co/flyingfishinwater/good_and_small_models/resolve/main/LFM2-1.2B-Q4_0.gguf](https://huggingface.co/flyingfishinwater/good_and_small_models/resolve/main/LFM2-1.2B-Q4_0.gguf)  

**Model Info URL:** [https://huggingface.co/LiquidAI/LFM2-1.2B](https://huggingface.co/LiquidAI/LFM2-1.2B)  

**Model License:** [License Info](https://huggingface.co/LiquidAI/LFM2-1.2B/raw/main/LICENSE)  

**Model Description:** LFM2 is a new generation of hybrid models developed by Liquid AI, specifically designed for edge AI and on-device deployment. It sets a new standard in terms of quality, speed, and memory efficiency. LFM2 is a new hybrid Liquid model with multiplicative gates and short convolutions. It supported languages: English, Arabic, Chinese, French, German, Japanese, Korean, and Spanish.  

**Developer:** [https://huggingface.co/LiquidAI](https://huggingface.co/LiquidAI)  

**File Size:** 696 MB  

**Context Length:** 4096 tokens  

**Prompt Format:** 
```

```

**Template Name:** qwen  

**Add BOS Token:** Yes  

**Add EOS Token:** No  

**Parse Special Tokens:** Yes  


---

## Jan v1 4B
Jan-v1-4B is an advanced agentic language model with 4.02 billion parameters, built on Qwen3-4B-Thinking. It is specifically designed for agentic reasoning and problem-solving, optimized for integration with Jan App. The model achieves strong performance on chat and question-answering benchmarks with improved reasoning capabilities, making it ideal for complex task automation and intelligent agent applications.  

**Model Intention:** Advanced agentic language model optimized for reasoning and problem-solving with 91.1% accuracy on question answering  

**Model URL:** [https://huggingface.co/flyingfishinwater/good_and_small_models/resolve/main/Jan-v1-4B-Q4_0.gguf](https://huggingface.co/flyingfishinwater/good_and_small_models/resolve/main/Jan-v1-4B-Q4_0.gguf)  

**Model Info URL:** [https://huggingface.co/janhq/Jan-v1-4B](https://huggingface.co/janhq/Jan-v1-4B)  

**Model License:** [License Info](https://www.apache.org/licenses/LICENSE-2.0.txt)  

**Model Description:** Jan-v1-4B is an advanced agentic language model with 4.02 billion parameters, built on Qwen3-4B-Thinking. It is specifically designed for agentic reasoning and problem-solving, optimized for integration with Jan App. The model achieves strong performance on chat and question-answering benchmarks with improved reasoning capabilities, making it ideal for complex task automation and intelligent agent applications.  

**Developer:** [https://huggingface.co/janhq](https://huggingface.co/janhq)  

**File Size:** 2400 MB  

**Context Length:** 2048 tokens  

**Prompt Format:** 
```

```

**Template Name:** qwen  

**Add BOS Token:** Yes  

**Add EOS Token:** No  

**Parse Special Tokens:** Yes  


---

## Menlo Lucy 1.7B
Lucy is a compact but capable 1.7B model focused on agentic web search and lightweight browsing. It is built on Qwen3-1.7B and optimized to run efficiently on mobile devices, even with CPU-only configurations. It was developed by Alan Dao, Bach Vu Dinh, Alex Nguyen, and Norapat Buppodom.  

**Model Intention:** Lucy is a compact but capable 1.7B model focused on agentic web search and lightweight browsing.  

**Model URL:** [https://huggingface.co/flyingfishinwater/good_and_small_models/resolve/main/Menlo_Lucy-Q4_K_M.gguf](https://huggingface.co/flyingfishinwater/good_and_small_models/resolve/main/Menlo_Lucy-Q4_K_M.gguf)  

**Model Info URL:** [https://huggingface.co/Menlo/Lucy](https://huggingface.co/Menlo/Lucy)  

**Model License:** [License Info](https://www.apache.org/licenses/LICENSE-2.0.txt)  

**Model Description:** Lucy is a compact but capable 1.7B model focused on agentic web search and lightweight browsing. It is built on Qwen3-1.7B and optimized to run efficiently on mobile devices, even with CPU-only configurations. It was developed by Alan Dao, Bach Vu Dinh, Alex Nguyen, and Norapat Buppodom.  

**Developer:** [https://huggingface.co/Menlo](https://huggingface.co/Menlo)  

**File Size:** 1056 MB  

**Context Length:** 2048 tokens  

**Prompt Format:** 
```

```

**Template Name:** qwen  

**Add BOS Token:** Yes  

**Add EOS Token:** No  

**Parse Special Tokens:** Yes  


---

## Nemotron 1.5B
OpenReasoning-Nemotron-1.5B is a large language model (LLM) which is a derivative of Qwen2.5-1.5B-Instruct. It is a reasoning model that is post-trained for reasoning about math, code and science solution generation. This model is ready for commercial/non-commercial research use.  

**Model Intention:** It is a reasoning model that is post-trained for reasoning about math, code and science solution generation.  

**Model URL:** [https://huggingface.co/flyingfishinwater/good_and_small_models/resolve/main/OpenReasoning-Nemotron-1.5B-Q4_K_M.gguf](https://huggingface.co/flyingfishinwater/good_and_small_models/resolve/main/OpenReasoning-Nemotron-1.5B-Q4_K_M.gguf)  

**Model Info URL:** [https://huggingface.co/nvidia/OpenReasoning-Nemotron-1.5B](https://huggingface.co/nvidia/OpenReasoning-Nemotron-1.5B)  

**Model License:** [License Info](https://huggingface.co/datasets/choosealicense/licenses/raw/main/markdown/cc-by-4.0.md)  

**Model Description:** OpenReasoning-Nemotron-1.5B is a large language model (LLM) which is a derivative of Qwen2.5-1.5B-Instruct. It is a reasoning model that is post-trained for reasoning about math, code and science solution generation. This model is ready for commercial/non-commercial research use.  

**Developer:** [https://huggingface.co/nvidia](https://huggingface.co/nvidia)  

**File Size:** 940 MB  

**Context Length:** 2048 tokens  

**Prompt Format:** 
```

```

**Template Name:** qwen  

**Add BOS Token:** Yes  

**Add EOS Token:** No  

**Parse Special Tokens:** Yes  


---

## Qwen3 1.7B Uncensored
Qwen3 1.7B Uncensored is an unrestricted variant designed for creative writing and storytelling without content limitations. It excels at generating fiction stories, horror narratives, plot development, scene continuation, and roleplaying scenarios. This model provides unfiltered responses and can produce intense or graphic content, making it suitable for users seeking unrestricted AI interactions for creative purposes.  

**Model Intention:** An uncensored 1.7B model optimized for creative writing, fiction stories, horror narratives, and unrestricted conversational scenarios.  

**Model URL:** [https://huggingface.co/flyingfishinwater/good_and_small_models/resolve/main/Qwen3-1.7B-Uncensored.gguf](https://huggingface.co/flyingfishinwater/good_and_small_models/resolve/main/Qwen3-1.7B-Uncensored.gguf)  

**Model Info URL:** [https://huggingface.co/DavidAU/Qwen3-1.7B-HORROR-Imatrix-Max-GGUF](https://huggingface.co/DavidAU/Qwen3-1.7B-HORROR-Imatrix-Max-GGUF)  

**Model License:** [License Info](https://www.apache.org/licenses/LICENSE-2.0.txt)  

**Model Description:** Qwen3 1.7B Uncensored is an unrestricted variant designed for creative writing and storytelling without content limitations. It excels at generating fiction stories, horror narratives, plot development, scene continuation, and roleplaying scenarios. This model provides unfiltered responses and can produce intense or graphic content, making it suitable for users seeking unrestricted AI interactions for creative purposes.  

**Developer:** [https://huggingface.co/DavidAU](https://huggingface.co/DavidAU)  

**File Size:** 1110 MB  

**Context Length:** 2048 tokens  

**Prompt Format:** 
```

```

**Template Name:** qwen  

**Add BOS Token:** Yes  

**Add EOS Token:** No  

**Parse Special Tokens:** Yes  


---

## Gemma 3 270M
Gemma 3 270M is an ultra-compact transformer model with 268M parameters, designed for efficient deployment on mobile and edge devices. Part of Google's Gemma family, it offers strong performance for its size with 32K context length, multilingual support, and responsible AI design. Ideal for applications requiring fast inference with minimal computational resources while maintaining quality text generation capabilities.  

**Model Intention:** Ultra-compact 270M parameter model optimized for resource-constrained environments with 32K context length  

**Model URL:** [https://huggingface.co/flyingfishinwater/good_and_small_models/resolve/main/gemma-3-270m-q4_0.gguf](https://huggingface.co/flyingfishinwater/good_and_small_models/resolve/main/gemma-3-270m-q4_0.gguf)  

**Model Info URL:** [https://huggingface.co/google/gemma-3-270m](https://huggingface.co/google/gemma-3-270m)  

**Model License:** [License Info](https://ai.google.dev/gemma/terms)  

**Model Description:** Gemma 3 270M is an ultra-compact transformer model with 268M parameters, designed for efficient deployment on mobile and edge devices. Part of Google's Gemma family, it offers strong performance for its size with 32K context length, multilingual support, and responsible AI design. Ideal for applications requiring fast inference with minimal computational resources while maintaining quality text generation capabilities.  

**Developer:** [https://huggingface.co/google](https://huggingface.co/google)  

**File Size:** 245 MB  

**Context Length:** 4096 tokens  

**Prompt Format:** 
```

```

**Template Name:** gemma  

**Add BOS Token:** Yes  

**Add EOS Token:** No  

**Parse Special Tokens:** Yes  


---

## LFM2 2.6B
LFM2-2.6B is a next-generation hybrid model by Liquid AI with 2.6B parameters, designed for edge AI and on-device deployment. It features multiplicative gates and short convolutions, offering 3x faster training and 2x faster decode/prefill speed on CPU. The model excels at agentic tasks, data extraction, RAG, creative writing, and multi-turn conversations. It supports 8 languages (English, Arabic, Chinese, French, German, Japanese, Korean, Spanish) with 32,768 context length and runs efficiently on CPU, GPU, and NPU hardware.  

**Model Intention:** Advanced hybrid model with 3x faster training and 2x faster inference, optimized for agentic tasks, RAG, and multi-turn conversations  

**Model URL:** [https://huggingface.co/flyingfishinwater/good_and_small_models/resolve/main/LFM2-2.6B-Q4_0.gguf](https://huggingface.co/flyingfishinwater/good_and_small_models/resolve/main/LFM2-2.6B-Q4_0.gguf)  

**Model Info URL:** [https://huggingface.co/LiquidAI/LFM2-2.6B](https://huggingface.co/LiquidAI/LFM2-2.6B)  

**Model License:** [License Info](https://huggingface.co/LiquidAI/LFM2-2.6B/raw/main/LICENSE)  

**Model Description:** LFM2-2.6B is a next-generation hybrid model by Liquid AI with 2.6B parameters, designed for edge AI and on-device deployment. It features multiplicative gates and short convolutions, offering 3x faster training and 2x faster decode/prefill speed on CPU. The model excels at agentic tasks, data extraction, RAG, creative writing, and multi-turn conversations. It supports 8 languages (English, Arabic, Chinese, French, German, Japanese, Korean, Spanish) with 32,768 context length and runs efficiently on CPU, GPU, and NPU hardware.  

**Developer:** [https://huggingface.co/LiquidAI](https://huggingface.co/LiquidAI)  

**File Size:** 1500 MB  

**Context Length:** 2048 tokens  

**Prompt Format:** 
```

```

**Template Name:** chatml  

**Add BOS Token:** Yes  

**Add EOS Token:** No  

**Parse Special Tokens:** Yes  


---

## LFM2-VL 1.6B
LFM2-VL-1.6B is an advanced multimodal vision-language model by Liquid AI featuring a 1.3B language model with 297M vision encoder. It processes images up to 512×512 pixels with variable resolutions, offers fast inference speed with superior performance compared to the 450M version, and supports 32,768 context length. Optimized for edge AI deployment with hybrid conv+attention architecture and SigLIP2 NaFlex vision encoder, providing enhanced reasoning and understanding capabilities.  

**Model Intention:** Enhanced multimodal vision-language model with improved reasoning capabilities, optimized for edge AI and low-latency applications  

**Model URL:** [https://huggingface.co/flyingfishinwater/good_and_small_models/resolve/main/LFM2-VL-1.6B-Q4_0.gguf](https://huggingface.co/flyingfishinwater/good_and_small_models/resolve/main/LFM2-VL-1.6B-Q4_0.gguf)  

**Model Info URL:** [https://huggingface.co/LiquidAI/LFM2-VL-1.6B](https://huggingface.co/LiquidAI/LFM2-VL-1.6B)  

**Model License:** [License Info](https://huggingface.co/LiquidAI/LFM2-VL-1.6B/raw/main/LICENSE)  

**Model Description:** LFM2-VL-1.6B is an advanced multimodal vision-language model by Liquid AI featuring a 1.3B language model with 297M vision encoder. It processes images up to 512×512 pixels with variable resolutions, offers fast inference speed with superior performance compared to the 450M version, and supports 32,768 context length. Optimized for edge AI deployment with hybrid conv+attention architecture and SigLIP2 NaFlex vision encoder, providing enhanced reasoning and understanding capabilities.  

**Developer:** [https://huggingface.co/LiquidAI](https://huggingface.co/LiquidAI)  

**File Size:** 900 MB  

**Context Length:** 4096 tokens  

**Prompt Format:** 
```

```

**Template Name:** chatml  

**Add BOS Token:** Yes  

**Add EOS Token:** No  

**Parse Special Tokens:** Yes  


---

## Qwen2.5-VL 3B Instruct
Qwen2.5-VL-3B-Instruct is a multimodal vision-language model with 3.09B parameters, featuring enhanced capabilities in coding, mathematics, and instruction following. It supports 29+ languages with up to 128K context length and 8K generation tokens. The model uses transformer architecture with RoPE, SwiGLU, and RMSNorm, offering improved resilience to diverse system prompts and specialized structured data understanding.  

**Model Intention:** Multimodal vision-language model with enhanced instruction following, coding, mathematics, and multilingual capabilities up to 128K context  

**Model URL:** [https://huggingface.co/flyingfishinwater/good_and_small_models/resolve/main/Qwen2.5-VL-3B-Instruct-Q4_K_M.gguf](https://huggingface.co/flyingfishinwater/good_and_small_models/resolve/main/Qwen2.5-VL-3B-Instruct-Q4_K_M.gguf)  

**Model Info URL:** [https://huggingface.co/Qwen/Qwen2.5-3B-Instruct-GGUF](https://huggingface.co/Qwen/Qwen2.5-3B-Instruct-GGUF)  

**Model License:** [License Info](https://huggingface.co/Qwen/Qwen2.5-3B-Instruct/raw/main/LICENSE)  

**Model Description:** Qwen2.5-VL-3B-Instruct is a multimodal vision-language model with 3.09B parameters, featuring enhanced capabilities in coding, mathematics, and instruction following. It supports 29+ languages with up to 128K context length and 8K generation tokens. The model uses transformer architecture with RoPE, SwiGLU, and RMSNorm, offering improved resilience to diverse system prompts and specialized structured data understanding.  

**Developer:** [https://huggingface.co/Qwen](https://huggingface.co/Qwen)  

**File Size:** 1930 MB  

**Context Length:** 2048 tokens  

**Prompt Format:** 
```

```

**Template Name:** qwen  

**Add BOS Token:** Yes  

**Add EOS Token:** No  

**Parse Special Tokens:** Yes  


---

## Qwen3-VL 4B Instruct
Qwen3-VL-4B-Instruct is a multimodal vision-language model with 4B parameters, featuring enhanced capabilities in instruction following, coding, mathematics, and multilingual understanding. It supports both image and text processing with strong reasoning capabilities, making it ideal for applications requiring visual understanding and text generation.  

**Model Intention:** Multimodal vision-language model with enhanced instruction following, coding, mathematics, and multilingual capabilities  

**Model URL:** [https://huggingface.co/flyingfishinwater/good_and_small_models/resolve/main/Qwen3-VL-4B-Instruct-Q4_0.gguf](https://huggingface.co/flyingfishinwater/good_and_small_models/resolve/main/Qwen3-VL-4B-Instruct-Q4_0.gguf)  

**Model Info URL:** [https://huggingface.co/Qwen/Qwen3-VL-4B-Instruct](https://huggingface.co/Qwen/Qwen3-VL-4B-Instruct)  

**Model License:** [License Info](https://www.apache.org/licenses/LICENSE-2.0.txt)  

**Model Description:** Qwen3-VL-4B-Instruct is a multimodal vision-language model with 4B parameters, featuring enhanced capabilities in instruction following, coding, mathematics, and multilingual understanding. It supports both image and text processing with strong reasoning capabilities, making it ideal for applications requiring visual understanding and text generation.  

**Developer:** [https://huggingface.co/Qwen](https://huggingface.co/Qwen)  

**File Size:** 2400 MB  

**Context Length:** 2048 tokens  

**Prompt Format:** 
```

```

**Template Name:** qwen  

**Add BOS Token:** Yes  

**Add EOS Token:** No  

**Parse Special Tokens:** Yes  


---

## Qwen3-VL 4B Thinking
Qwen3-VL-4B-Thinking is a specialized multimodal vision-language model with enhanced reasoning capabilities and thinking mode. It excels at complex visual reasoning tasks including mathematical problem solving, scientific analysis, coding with visual inputs, and intricate logical reasoning. The thinking mode enables step-by-step problem solving with both images and text, making it ideal for applications requiring deep analytical capabilities and visual understanding.  

**Model Intention:** Advanced multimodal reasoning model with thinking mode for complex visual reasoning, mathematics, and scientific tasks  

**Model URL:** [https://huggingface.co/flyingfishinwater/good_and_small_models/resolve/main/Qwen3-VL-4B-Thinking-Q4_0.gguf](https://huggingface.co/flyingfishinwater/good_and_small_models/resolve/main/Qwen3-VL-4B-Thinking-Q4_0.gguf)  

**Model Info URL:** [https://huggingface.co/Qwen/Qwen3-VL-4B-Thinking](https://huggingface.co/Qwen/Qwen3-VL-4B-Thinking)  

**Model License:** [License Info](https://www.apache.org/licenses/LICENSE-2.0.txt)  

**Model Description:** Qwen3-VL-4B-Thinking is a specialized multimodal vision-language model with enhanced reasoning capabilities and thinking mode. It excels at complex visual reasoning tasks including mathematical problem solving, scientific analysis, coding with visual inputs, and intricate logical reasoning. The thinking mode enables step-by-step problem solving with both images and text, making it ideal for applications requiring deep analytical capabilities and visual understanding.  

**Developer:** [https://huggingface.co/Qwen](https://huggingface.co/Qwen)  

**File Size:** 2100 MB  

**Context Length:** 2048 tokens  

**Prompt Format:** 
```

```

**Template Name:** qwen  

**Add BOS Token:** Yes  

**Add EOS Token:** No  

**Parse Special Tokens:** Yes  


---

## Qwen3-VL 2B Instruct
Qwen3-VL-2B-Instruct is a compact multimodal vision-language model with 2B parameters, designed for efficient deployment while maintaining strong performance in visual understanding and text generation. It supports both image and text processing with enhanced instruction following capabilities, making it ideal for applications requiring visual understanding with resource constraints. The model offers multilingual support and robust reasoning capabilities.  

**Model Intention:** Compact multimodal vision-language model with enhanced instruction following, optimized for efficient deployment  

**Model URL:** [https://huggingface.co/flyingfishinwater/good_and_small_models/resolve/main/Qwen3-VL-2B-Instruct-Q4_0.gguf](https://huggingface.co/flyingfishinwater/good_and_small_models/resolve/main/Qwen3-VL-2B-Instruct-Q4_0.gguf)  

**Model Info URL:** [https://huggingface.co/Qwen/Qwen3-VL-2B-Instruct](https://huggingface.co/Qwen/Qwen3-VL-2B-Instruct)  

**Model License:** [License Info](https://www.apache.org/licenses/LICENSE-2.0.txt)  

**Model Description:** Qwen3-VL-2B-Instruct is a compact multimodal vision-language model with 2B parameters, designed for efficient deployment while maintaining strong performance in visual understanding and text generation. It supports both image and text processing with enhanced instruction following capabilities, making it ideal for applications requiring visual understanding with resource constraints. The model offers multilingual support and robust reasoning capabilities.  

**Developer:** [https://huggingface.co/Qwen](https://huggingface.co/Qwen)  

**File Size:** 1300 MB  

**Context Length:** 2048 tokens  

**Prompt Format:** 
```

```

**Template Name:** qwen  

**Add BOS Token:** Yes  

**Add EOS Token:** No  

**Parse Special Tokens:** Yes  


---

## Ministral 3 3B Instruct 2512
Ministral-3-3B-Instruct-2512 is a multimodal vision-language model with 3B parameters, designed for efficient deployment while maintaining strong performance in visual understanding and text generation. It supports both image and text processing with enhanced instruction following capabilities, making it ideal for applications requiring visual understanding with resource constraints. The model offers multilingual support and robust reasoning capabilities.  

**Model Intention:** Multimodal vision-language model with enhanced instruction following, optimized for efficient deployment and visual understanding  

**Model URL:** [https://huggingface.co/flyingfishinwater/good_and_small_models/resolve/main/Ministral-3-3B-Instruct-2512-Q4_0.gguf](https://huggingface.co/flyingfishinwater/good_and_small_models/resolve/main/Ministral-3-3B-Instruct-2512-Q4_0.gguf)  

**Model Info URL:** [https://huggingface.co/mistralai/Ministral-3-3B-Instruct-2512](https://huggingface.co/mistralai/Ministral-3-3B-Instruct-2512)  

**Model License:** [License Info](https://www.apache.org/licenses/LICENSE-2.0.txt)  

**Model Description:** Ministral-3-3B-Instruct-2512 is a multimodal vision-language model with 3B parameters, designed for efficient deployment while maintaining strong performance in visual understanding and text generation. It supports both image and text processing with enhanced instruction following capabilities, making it ideal for applications requiring visual understanding with resource constraints. The model offers multilingual support and robust reasoning capabilities.  

**Developer:** [https://huggingface.co/mistralai](https://huggingface.co/mistralai)  

**File Size:** 1900 MB  

**Context Length:** 4096 tokens  

**Prompt Format:** 
```

```

**Template Name:** chatml  

**Add BOS Token:** Yes  

**Add EOS Token:** No  

**Parse Special Tokens:** Yes  


---