GitHub seismograph — discoveries and growing interest
Explore models for text, images, video and audio from Hugging Face’s trending top 50. Quantized versions of the same model are grouped together.
Measured: · Rank changes since 13.10.2026 18:57 UTC
Models shown: 40 / 40
Open multimodal embedding model from Google DeepMind that maps text, code, images, video, and audio into a unified 768-dimensional vector space. Runs on mobile devices and laptops for search, RAG, classification, and clustering.
Embeddings · Created 2026-09-14 · apache-2.0
A local 12B model that rewrites AI-written drafts (emails, essays, reports, forum posts; English and Chinese) so they read as if a person wrote them, preserving numbers, dates, names and quotes.
Text and chat · Created 2026-09-11 · apache-2.0
An accelerated Qwen-Image-2.1 checkpoint for text-to-image generation and image editing in just 8 denoising steps. It loads via QwenImage21Pipeline in Diffusers without manual scheduler configuration.
Image generation · Created 2026-10-09 · other
Quantized builds of the Qwen-Image-2.1 text-to-image model for local generation in ComfyUI, including uncensored variants without a built-in content filter. The repo ships GGUF, Safetensors and MLX files plus companion text encoder and VAE.
Image generation · Version created 2026-09-20 · other
MiniCPM-V 4.7 in the 35B-A3B variant is the first open-source MoE model in the series, handling single-image, multi-image and video understanding and designed for on-device inference. With Lookahead Expert Scheduling, layer-wise quantization and speculative sampling it runs on iPhone at 13+ tokens/s using just 3 GB of memory.
Vision and chat · Created 2026-10-07 · apache-2.0
A 7.89 GB GGUF quantization of Qwen3.8-27B for stock llama.cpp that keeps tool calling intact and runs on modest hardware. An optional vision add-on enables image input.
Text and chat · Created 2026-10-08 · apache-2.0
A 27B multimodal model that takes a state (text, JSON, images, video) plus a schema of typed questions and returns probabilities for every allowed option in a single forward pass, with no free-form text generation.
Vision and chat · Created 2026-09-30 · apache-2.0
A 3B-parameter multimodal decision model built on LFM2.5-VL-3B. It takes a state (text, JSON, images) and a set of questions, returning calibrated typed answers in one forward pass with zero output tokens.
Vision and chat · Created 2026-10-05 · other
GGUF builds of the Qwen3-VL-8B-Instruct text encoder with its refusal direction projected out, for running Qwen-Image 2.1 Turbo locally via stable-diffusion.cpp. Includes quantized encoder variants and mmproj.
Image generation · Version created 2026-10-09 · apache-2.0
Open vision-language model Qwen3.8-27B (27B params, up to 1M token context) for coding, agentic tasks, and image/video understanding. Weights in Hugging Face Transformers format, compatible with vLLM, SGLang, and TokenSpeed.
Vision and chat · Created 2026-08-05 · apache-2.0
Turkish text-to-speech model: a 5.6M-parameter DiT acoustic model plus a 3M-parameter vocoder, 8.6M total (~34 MB), Apache 2.0. Runs fully offline on GPU or CPU, supports streaming generation and batched calls.
Speech synthesis · Created 2026-10-01 · apache-2.0
A post-trained multimodal model based on Qwen3.8-27B that fixes the habit of endlessly re-deriving an answer until hitting the context cap. The repo ships weights in many precisions (BF16, FP8, NVFP4, INT8/INT4, GGUF) with SGLang, vLLM, llama.cpp and NInfer support.
Vision and chat · Created 2026-10-04 · apache-2.0
Open-weights 35B-parameter decision model (about 3B active per token) built on Qwen3.5-35B-A3B, designed to pick one calibrated answer from up to 256 options in a single pass for ranking, classification and search re-ranking.
Other tasks · Created 2026-10-11 · apache-2.0
Baidu's OCR model for text recognition on images and multi-page PDFs, handling long documents in a single pass. Runs via Transformers, vLLM, or SGLang.
Vision and chat · Created 2026-06-19 · mit
Vision and chat
Vision and chat · Version created 2026-09-01 · apache-2.0
Vision and chat
Vision and chat · Version created 2026-09-07 · apache-2.0
Vision and chat
Vision and chat · Created 2026-09-10 · mit
Speech recognition
Speech recognition · Created 2025-04-15 · cc-by-4.0 · Access requires approval
Vision and chat
Vision and chat · Version created 2026-10-09 · mit
Text and chat
Text and chat · Created 2026-09-20 · apache-2.0
Vision and chat
Vision and chat · Version created 2026-10-02 · apache-2.0