Fathom-DeepResearch is a dual-model intelligent system composed of Fathom-Search-4B and Fathom-Synthesizer-4B. It is optimized for long-term network search, evidence verification, and generating reports rich in references. This system achieves the optimal performance of open-source weights in search-intensive tasks, surpassing multiple closed-source deep research agents.

Natural Language ProcessingTransformersEnglish

This is a quantized version of Tesslate/WEBGEN-4B-Preview, focusing on generating single-file websites and capable of transforming prompts into concise, responsive HTML/CSS/Tailwind code. The model is small and suitable for local operation and rapid iteration, with features such as open weights, mobile-first output, and no external JS by default.

Natural Language ProcessingTransformersEnglish

FLUX.1 Krea [dev] is a rectified flow transformer with 12 billion parameters, capable of generating high-quality images based on text descriptions. This model focuses on aesthetic photography and excels in image quality and instruction following. It uses guided distillation for efficient training and opens its weights to promote scientific research and artistic creation.

MultimodalDiffusersEnglish

FLUX.1 Krea [dev] is an advanced rectified flow transformer model with 12 billion parameters, specifically designed to generate high-quality images based on text descriptions. This model excels in the field of aesthetic photography, has excellent prompt-following ability, is trained using guided distillation technology, and its weights are open for non-commercial use.

Computer VisionDiffusersEnglish

FLUX.1 Krea [dev] is a rectified flow transformer with 12 billion parameters, specifically designed to generate high-quality images based on text descriptions. It excels in output quality, instruction following, etc., focuses on aesthetic photography, adopts an efficient guided distillation training method, and opens its weights to promote scientific research and creative work.

Computer VisionDiffusersEnglish

DeepSeek-R1T-Chimera is an open-source weights model that combines the intelligence of DeepSeek-R1 with the token efficiency of DeepSeek-V3.

Natural Language ProcessingTransformers

This is the FP8 quantized version of the OpenAI whisper-large-v3 model. It uses the FP8 data type to quantize the model weights and activations, enabling efficient inference through vLLM. It is specifically designed for speech recognition tasks, capable of processing audio input and outputting text content.

Audio ProcessingTransformersEnglish

This is a quantized version of the OpenAI whisper-large-v3 model. By quantizing the weights to the INT4 data type, the inference efficiency is effectively improved. This model is specifically designed for speech recognition tasks, supports multilingual speech-to-text functionality, and is suitable for scenarios requiring efficient speech processing.

Audio ProcessingTransformersEnglish

Cohere Labs Command A is a research version of a 111-billion parameter model with open weights, optimized for the demanding enterprise needs that require fast, secure, and high-quality AI.

Natural Language ProcessingTransformersMultiple Languages

This is the INT8 quantized version of the IBM granite-3.1-8b-instruct model. It improves inference efficiency by quantizing weights and activations to the INT8 data type and is suitable for text generation tasks. It achieved an average score of 70.26 in the OpenLLM benchmark test, close to the 70.30 score of the original model.

Natural Language ProcessingSafetensorsEnglish

The EVA CLIP model is a vision-language model based on OpenCLIP and timm model weights, supporting tasks such as zero-shot image classification.

Computer VisionSafetensors

whisper-large-v3-int4-ov is an automatic speech recognition model that converts OpenAI's whisper-large-v3 model into OpenVINO IR format and quantizes it to INT4 weights. It supports multilingual speech transcription and has efficient inference performance.

Audio ProcessingOpenvinoMultiple Languages

This is a converted and quantized speech recognition model. It is converted from the distil-large-v3 model to the OpenVINO™ IR format, and the weights are compressed to INT8 to improve performance and compatibility.

Audio ProcessingTransformersEnglish

Command R+ is a research version model with open weights and 104 billion parameters launched by Cohere Labs. It has retrieval-augmented generation (RAG) and tool usage capabilities, and supports multilingual and multi-step task automation.

Natural Language ProcessingTransformersMultiple Languages

C4AI Command-R is a high-performance generative model research version with 35 billion parameters, using the GGUF format. It is a large language model with open weights, optimized for various use cases such as inference, summarization, and question-answering, and has multilingual generation capabilities and high-performance RAG capabilities.

Natural Language ProcessingGguf

OTTER-MPT7B-Init is a set of weights for initializing Otter model training, converted directly from Openflamingo.

MultimodalTransformers

A text-to-image model fine-tuned from Stable Diffusion v1-5, developed independently by Predogl and piEsposito, with open weights and configurations supporting high-quality image generation from concise prompts

Computer VisionDiffusers

LoRA adaptation weights trained on Stable Diffusion 2.1-base, optimized for the Openjourney model, provided by PromptHero

Computer VisionDiffusers

A text-to-image model fine-tuned from Stable Diffusion v1-5, trained by independent developers Predogl and piEsposito, with open weights and configurations

Computer VisionDiffusers