Qwen-Image is a text-to-image generation model developed based on the Qwen series. It supports both Chinese and English inputs and performs excellently on multiple evaluation metrics. It is particularly suitable for image generation scenarios that pursue realistic effects.

MultimodalDiffusersMultiple Languages

This is a video understanding model based on the VideoMAE architecture. It was pre-trained on the Kinetics dataset and fine-tuned on an unknown dataset that may be related to sign language recognition. The model achieved an accuracy of 78.11% on the evaluation set and is suitable for video classification tasks.

Computer VisionTransformers

Z-Image is an efficient basic image generation model with 6 billion parameters, specifically designed to address the efficiency and quality issues in the field of image generation. Its distilled version, Z-Image-Turbo, can reach or exceed the leading competitors with only 8 function evaluations. It can achieve sub-second inference latency on enterprise-level H800 GPUs and run on consumer-grade devices with 16G VRAM.

Computer VisionDiffusersMultiple Languages

This is a video understanding model based on the VideoMAE architecture. It has been fine-tuned on the basis of pre-training on the Kinetics dataset and is specifically designed for sign language recognition tasks. The model's performance on the evaluation set needs improvement, with an accuracy of 0.0010.

Computer VisionTransformers

This is a video understanding model based on the VideoMAE-base architecture, which has been fine-tuned for 20 epochs on an unknown dataset. The model's performance on the evaluation set is limited, with an accuracy of 0.0041 and a loss value of 7.7839.

Computer VisionTransformers

This model is a joke rating model fine-tuned on multilingual text based on FacebookAI/xlm-roberta-large. It is specifically designed to evaluate the quality and humor level of jokes. It achieved an accuracy of 0.4005 and a root mean square error of 5.0327 on the evaluation set.

Natural Language ProcessingTransformers

This is a video understanding model fine-tuned on an unknown dataset based on the MCG-NJU/videomae-base model. After 20 epochs of training, it achieved an accuracy of 13.31% on the evaluation set. This model is specifically optimized for video analysis tasks.

Computer VisionTransformers

This model is a large language model fine-tuned on a specific dataset based on meta-llama/Llama-3.1-70B-Instruct. It is specifically designed for text generation tasks and achieved a loss value of 0.6542 on the evaluation set.

Natural Language ProcessingPeft

This is an English-Spanish translation model fine-tuned on the Amazon dataset based on the Google mT5-small model, specifically designed for text abstract generation tasks. The model achieved scores of Rouge1: 16.44 and Rouge2: 8.04 on the evaluation set.

Natural Language ProcessingTransformers

The BT4 model is the neural network model behind the LeelaChessZero engine, specifically designed for chess games. This model is based on the Transformer architecture and can predict the best next move based on historical moves, evaluate the game situation, and generate move probabilities.

Scientific ComputingTransformers

This is a video action recognition model fine-tuned on the WLASL dataset based on the VideoMAE-base architecture, specifically optimized for sign language recognition tasks, achieving an accuracy of 48.22% on the evaluation set.

Computer VisionTransformers

This is a video action recognition model fine-tuned on the WLASL dataset based on the VideoMAE-Base architecture. After 200 epochs of training, it achieved a top-1 accuracy of 52.96% and a top-5 accuracy of 79.88% on the evaluation set, specifically designed for sign language action recognition tasks.

Computer VisionTransformers

This model is a professional code generation model fine-tuned on the swe_only_sweagent dataset based on Qwen2.5-Coder-32B-Instruct. It achieved a loss value of 0.1210 on the evaluation set and is specifically optimized for software engineering-related tasks.

Natural Language ProcessingTransformers

metatune-gpt20b is a large language model prototype with self-improvement ability. It can generate new data for itself, evaluate its own performance, and adjust hyperparameters according to improvement indicators. This model performs excellently in scientific and mathematical understanding at the postdoctoral level and can also be used for coding tasks.

Natural Language ProcessingTransformersEnglish

This is a quantized version of unsloth/Mistral-Small-3.2-24B-Instruct-2506. By quantizing the weights and activation functions to the FP4 data type, it reduces the disk size and GPU memory requirements while supporting vLLM inference. It has been evaluated on multiple tasks to compare the quality with the non-quantized model.

Natural Language ProcessingSafetensorsMultiple Languages

This is a Vietnamese text correction model fine-tuned based on vinai/bartpho-syllable, which has achieved a good result of 69.12% character accuracy on the evaluation set and is specifically used for the automatic correction task of Vietnamese texts.

Natural Language ProcessingTransformers

This is a classifier model fine-tuned based on the jhu-clsp/mmBERT-small model, achieving an accuracy of 91.07% on the evaluation set. It is mainly used for text classification tasks.

Natural Language ProcessingPeft

G²RPO is a novel reinforcement learning framework specifically designed for preference alignment in flow models, significantly improving the generation quality through a granular reward evaluation mechanism.

MultimodalDiffusers

Qwen3-4B-SafeRL is a safety-aligned version based on the Qwen3-4B model. It is trained through reinforcement learning and combined with the reward signals of Qwen3Guard-Gen, enhancing the model's robustness against harmful or adversarial prompts. While ensuring safety, it avoids overly simple or evasive rejection behaviors.

Natural Language ProcessingTransformers

This is a text classification model fine-tuned based on the mmBERT-small architecture, specifically designed for the message type classification task. It achieved an accuracy of 93.94% on the evaluation set and has efficient text classification capabilities.

Natural Language ProcessingPeft