We're sharing/showcasing best of @huggingface models. Follow to stay in loop. Promoting Open-Source models.
Joined December 2025
- Tweets31.7K
- Following32
- Followers71.7K
- Likes620
With 214 downloads and growing, this model proves that powerful multimodal AI can be compact and locally runnable. It's a game changer for developers in Southeast Asia who need accurate, multilingual vision-language capabilities without cloud dependency.
4B parameters, vision-language architecture, quantized to GGUF for efficient inference on consumer hardware. Trained on diverse multilingual data covering English and 7 Southeast Asian languages, ensuring strong regional performance.
Build multilingual image captioning, visual question answering, or document understanding for Southeast Asian markets. This model understands both images and text across 8 regional languages, perfect for localized apps and accessibility tools.
Meet Qwen-SEA-LION-v4-4B-VL-GGUF: a 4B parameter vision-language model optimized for Southeast Asia. It speaks English, Vietnamese, Indonesian, Thai, Malay, Tamil, Tagalog, and Burmese, all in a GGUF package for easy local deployment.
With 173 downloads, it's a niche but valuable resource for Vietnamese math AI. Quantized GGUF makes it accessible on modest hardware. No likes yet, but it's a solid starting point for math-focused LLM projects in Vietnamese.
Based on BloomZ-7B1, a 7B parameter model with 30 layers and 4096 hidden size. Fine-tuned on Vietnamese math data. The GGUF quantization shrinks it for faster inference while keeping performance. Tags show English and Vietnamese support.
Use it to build Vietnamese math solvers, step-by-step tutors, or homework helpers. The GGUF format means you can run it locally with llama.cpp or similar tools. Great for education apps or research in low-resource math NLP.
Meet BloomZ-7B1-Vi-Math-GGUF: a quantized version of a 7B model fine-tuned for Vietnamese math. It's ready for transformers and runs efficiently on consumer hardware. Perfect for building math tutoring tools or solving problems in Vietnamese.
With 168 downloads and growing, this model proves that small can be smart. It delivers strong math reasoning in a compact, locally runnable package. Try it for your next project.
Built on DeepSeek-R1-14B, this GGUF version is quantized for efficiency. It's a base model fine-tuned for math reasoning, with a conversational interface and MIT license.
Stuck on a math problem? This model excels at chain-of-thought reasoning. Use it as a conversational tutor, a homework helper, or to build apps that explain solutions step by step.
Meet DeepSeek-R1-14B-CoT-Math-Reasoning-Full-GGUF: a 14B model that thinks step by step for math. Quantized for GGUF, it's ready to run locally and tackle your toughest problems.
With 543 downloads and growing, Pocket TTS Loom is a reliable choice for local TTS. Its compact size and fast inference make it ideal for real-time applications. Give it a try and let your projects speak!
Built on Kyutai's Pocket TTS and quantized to GGUF, this model is optimized for efficiency. The Loom runtime ensures smooth performance on edge devices. Licensed under CC-BY-4.0, so it's open for commercial use.
From audiobooks to voice assistants, Pocket TTS Loom turns text into natural speech. Build accessible apps, interactive toys, or voice-enabled interfaces. It's perfect for offline use cases where privacy and speed matter.
