LongCat-Video is a basic video generation model with 13.6 billion parameters, which performs excellently in generation tasks such as text-to-video, image-to-video, and video frame continuation. It is particularly good at efficient and high-quality long video generation.

MultimodalDiffusersMultiple Languages

CAST 0.7B is a speech-to-speech language model built on a 0.7B parameter Gemma3-style language model, capable of generating natural speech audio continuation content. This model relies on the CAST WavTokenizer for encoding/decoding operations and is specifically designed for speech generation tasks.

Audio ProcessingSafetensors

A lightweight large language model optimized for Japanese novel creation, based on the Mistral architecture, providing creative generation and text continuation functions

Natural Language ProcessingGgufJapanese

AuriStream-1B is a biologically inspired GPT-style autoregressive Transformer model specifically designed to predict cochlear markers in long speech contexts. The model uses a long context window of approximately 20 seconds (4096 tokens) and was trained on the LibriLight dataset (approximately 60,000 hours) for about 500,000 steps. It can learn rich time-aligned representations and generate speech continuations.

Audio ProcessingTransformersEnglish

This is a speech language model based on discrete Hubert tokens, focusing on efficient training and capable of generating speech segment continuations.

Audio ProcessingTransformers

A language model based on Qwen2.5-7B, continually pre-trained using LoRA technology on DTF Russian forum data

Natural Language ProcessingSafetensorsOther

A text classification model supporting continual learning and dynamic category expansion

Natural Language ProcessingSafetensorsOther

OpenHermes 2.5 Mistral 7B is a state-of-the-art model fine-tuned based on the Mistral architecture and is a continuation of OpenHermes 2, trained on an additional code dataset.

Natural Language ProcessingGguf

TiC-CLIP is a continually trained vision-language model focused on addressing the high cost of synchronizing base models with the latest data.

TiC-CLIP is a vision-language model trained on the TiC-DataComp-Yearly dataset, employing continual learning strategies to keep the model synchronized with the latest data

TiC-CLIP is an improved vision-language model based on OpenCLIP, focusing on temporal continual learning, with training data spanning from 2014 to 2022

TiC-CLIP is an improved vision-language model based on OpenCLIP, employing continual training strategies on time-series data to effectively reduce computational costs for model updates.

TiC-CLIP is a continually trained vision-language model that maintains synchronization with the latest data through temporally continuous training, avoiding the high costs of frequent retraining.

TiC-CLIP is an improved vision-language model based on OpenCLIP, focusing on continual temporal learning, with training data spanning from 2014 to 2022

MahaMarathi 7B is a domain-adapted continual pre-training instruction fine-tuned native Marathi large language model (LLM) based on the Llama2+Mistral architecture with 7 billion parameters, specifically designed for over 83 million native speakers, capable of fluently handling complex Marathi dialogues and instruction interactions.

Natural Language ProcessingTransformersOther

This model is a multilingual model fine-tuned on Swiss German text data through continual pre-training based on xlm-roberta-base, supporting both Swiss German and Standard German processing.

Natural Language ProcessingTransformersOther

A Vietnamese-optimized large language model based on Llama-2-7B, enhanced through continual pre-training on 124GB of multi-domain Vietnamese and English data for improved language understanding

Natural Language ProcessingTransformersMultiple Languages

A pre-trained continuation model based on Mistral-7B-v0.1, fine-tuned with 40GB of text data from the Swedish forum Flashback, supporting multilingual generation.

Natural Language ProcessingTransformersMultiple Languages

Generative AI-based MIDI music creation model supporting generation from scratch or template-based continuation

Audio ProcessingTransformers

A video classification model based on the TimeSformer architecture, pre-trained on the Kinetics-400 dataset and fine-tuned on the UCF101 dataset, utilizing LoRA technology for continual learning.

Computer VisionTransformers