Kimi Linear is a hybrid linear attention architecture that outperforms traditional full attention methods in various scenarios, including short and long contexts, as well as reinforcement learning extension mechanisms. It can effectively solve the problem of low efficiency of traditional attention mechanisms in long context tasks and bring more efficient solutions to fields such as natural language processing.

Natural Language ProcessingTransformers

A binary text classification model for detecting AI-generated content in Reddit comments, supporting real-time detection in browser extensions.

Natural Language ProcessingTransformers

Kanana 1.5 is a bilingual large language model developed by Kakao, with significant improvements in programming, mathematics, and function calling capabilities. It supports a 32K token context length and can process ultra-long texts up to 128K tokens using YaRN extension technology.

Natural Language ProcessingTransformersMultiple Languages

xGen-small is an enterprise-grade compact language model that achieves long-context performance with predictable low costs through domain-focused data curation, scalable pre-training, length extension, and reinforcement learning fine-tuning.

Natural Language ProcessingTransformersEnglish

Foundation-Sec-8B is an 8-billion-parameter foundational language model specifically designed for cybersecurity, based on the Llama-3.1-8B extension, suitable for security scenarios such as threat detection and vulnerability assessment.

Natural Language ProcessingTransformersEnglish

OctoThinker is a reinforcement learning - friendly base language model built on the Llama - 3 family of models, focusing on the research of reinforcement learning extension with mid - term training incentives. This model with a 3B parameter scale has important application value in the field of natural language processing and is particularly suitable for text generation tasks.

Natural Language ProcessingSafetensorsEnglish

A speech-language model based on Qwen2.5-7B extension, supporting speech-text interleaved training and cross-modal generation

MultimodalTransformersEnglish

A pose control extension developed based on the BRIA 2.3 text-to-image model, generating high-quality images through text prompts and human pose estimation from input images

Computer VisionDiffusers

Parler-TTS Mini Multilingual v1.1 is a multilingual extension based on the Parler-TTS Mini version, supporting text-to-speech in 8 European languages.

Audio ProcessingTransformersMultiple Languages

The Indic Parler-TTS Pretrained Model is a multilingual Indian language extension of Parler-TTS Mini, supporting 21 languages, including various Indian languages and English.

Audio ProcessingTransformersMultiple Languages

Indic Parler-TTS is a multilingual extension of Parler-TTS Mini, supporting 21 languages including various Indian languages and English.

Audio ProcessingTransformersMultiple Languages

EBEN is a bandwidth extension model for French speech, specifically designed to process audio collected by body-conducted voice sensors (such as throat microphones). It can reduce noise in low-frequency, noisy body-conducted voice and regenerate mid- and high-frequency components from low-frequency content, thereby significantly improving the clarity and quality of the audio. This model is trained on specific non-traditional sensor data and is suitable for voice enhancement tasks in specific fields.

Audio ProcessingTransformersFrench

A Tibetan vocabulary extension model based on the Llama2-7B architecture, enhancing Tibetan processing capabilities through continuous pre-training strategies

Natural Language ProcessingTransformers

A ControlNet extension developed based on the BRIA 2.3 text-to-image model, enabling color style control through color grids

Computer VisionDiffusers

A LoRA model for creating corporate headshots and portraits, especially suitable for LinkedIn profile photo facial extensions.

Computer VisionDiffusers

A ControlNet extension developed based on the BRIA 2.3 text-to-image model, enabling image color reset and diverse variant generation through text prompts and grayscale guidance

Computer VisionDiffusers

VidXTend is a streamlined encapsulation of StreamingT2V's second-stage video extension process, designed to extend 16-frame 256x256 pixel animations by 8 frames each time (equivalent to 1 second duration at 8fps)

Computer VisionDiffusers

Fluently V4-inpainting is a versatile image processing model created by merging multiple checkpoints and LoRA techniques, excelling particularly in image inpainting and extension tasks.

Computer VisionDiffusers

X-CLIP is a minimalist extension of CLIP for general video-language understanding, trained via contrastive learning to match videos and texts.

MultimodalTransformersEnglish

Yarn Mistral 7B 128K is an advanced language model optimized for long-context processing, further pre-trained on long-context data using the YaRN extension method, supporting a 128k token context window.

Natural Language ProcessingTransformersEnglish