CodeV is a 7-billion-parameter vision-language model fine-tuned based on Qwen2.5-VL-7B-Instruct. Through two-stage training of supervised fine-tuning (SFT) and reinforcement learning (RL) based on tool-aware policy optimization (TAPO), it aims to achieve reliable and interpretable visual reasoning. It represents visual tools as executable Python code and ensures the consistency between tool usage and question evidence through a reward mechanism, solving the problem of irrelevant tool invocation under high accuracy.

MultimodalTransformersMultiple Languages

Z-Image-Re-Turbo is a text-to-image generation model that has been optimized for de-reduction and re-acceleration based on the Z-Image-De-Turbo model. This model aims to balance the convenience during training and the speed during inference. It restores the fast generation ability close to the original Turbo model while maintaining the same training-friendly features as Z-Image-De-Turbo, enabling it to be perfectly compatible with a large number of trained LoRA models in the Z-Image ecosystem.

MultimodalDiffusers

This model is obtained through distillation training using the high-reasoning-difficulty Gemini 3 Pro preview dataset based on the Qwen3-4B-Thinking-2507 base model. It focuses on improving complex reasoning abilities in the fields of coding and science. Through training with specific datasets, it aims to efficiently transfer the reasoning abilities of large models (such as Gemini 3 Pro) to a smaller-scale model.

Natural Language ProcessingTransformers

VibeThinker-1.5B is a 1.5-billion-parameter dense language model launched by Weibo AI. It is fine-tuned based on Qwen2.5-Math-1.5B and is specifically designed for mathematical and algorithmic coding problems. Trained using the 'Spectrum to Signal Principle' framework, it outperforms larger models in multiple math competition tests. The training cost is approximately $7,800, and it supports an output of up to about 40k tokens.

Natural Language ProcessingTransformersMultiple Languages

Olmo 3 is a new generation of language model family developed by the Allen Institute for AI, including 7B and 32B instruction and thinking variants. This model performs excellently in long-chain thinking and can significantly improve the performance of reasoning tasks such as mathematics and coding. All code, checkpoints, and training details will be made public to promote the development of language model science.

Natural Language ProcessingTransformersEnglish

Olmo 3 is a series of language models developed by the Allen Institute for AI, including two scales of 7B and 32B, with two variants: instructional and reflective. This model performs excellently in long-chain thinking and can effectively improve the performance of reasoning tasks such as mathematics and coding. It adopts a multi-stage training method, including supervised fine-tuning, direct preference optimization, and reinforcement learning with verifiable rewards.

Natural Language ProcessingTransformersEnglish

Olmo-3-7B-Think-DPO is a 7B parameter language model developed by the Allen Institute for AI. It has the ability of long-chain thinking and performs excellently in reasoning tasks such as mathematics and coding. This model has undergone multi-stage training including supervised fine-tuning, direct preference optimization, and reinforcement learning based on verifiable rewards, and is designed specifically for research and educational purposes.

Natural Language ProcessingTransformersEnglish

xLSTM-7b-Instruct is an experimental fine-tuned version based on NX-AI/xLSTM-7b, specifically optimized for instruction-following tasks. This model adds support for chat templates and uses TRL for supervised fine-tuning training, aiming to provide a better conversational interaction experience.

Natural Language ProcessingTransformers

This is a text-to-video LoRA model trained based on the Wan2.2-T2V-A14B base model, using the AI Toolkit for training, specifically designed to enhance the text-to-video conversion ability.

MultimodalDiffusers

Apertus is a fully open multilingual large language model developed by Swiss AI, offering two parameter scales of 7 billion and 8 billion. This model supports over 1000 languages, uses fully compliant and open training data, and its performance can rival that of closed-source models. Apertus was pre-trained on 15T tokens and adopts a phased curriculum training method, supporting a context length of up to 65,536 tokens.

Natural Language ProcessingTransformersOther

Apertus is a fully open multilingual large language model developed by Swiss AI. It has parameter scales of 8 billion and 70 billion, supports over 1000 languages and long context processing, and only uses fully compliant open training data. Its performance can rival that of closed-source models.

Natural Language ProcessingTransformersOther

WEBGEN DEVSTRAL IMAGES is an AI model focused on web page generation. It can generate single-page web pages using HTML, CSS, JS, and Tailwind technologies. This project is trained based on custom templates and uses the supervised fine-tuning method, training with a dataset generated by GPT-OSS-120B.

Natural Language ProcessingTransformersEnglish

This model is based on GPT-OSS-20B and fine-tuned using the Unsloth reinforcement learning framework. The aim is to optimize inference efficiency and reduce vulnerabilities that occur during reinforcement learning from human feedback (RLHF) training. The fine-tuning process focuses on the robustness and efficiency of alignment, ensuring that the model maintains inference depth without incurring excessive computational overhead.

Natural Language ProcessingTransformersEnglish

TARS-SFT-7B is a security reasoning model based on supervised fine-tuning. It serves as the basic model for reinforcement learning training and is specifically designed to enhance the security of AI systems. This model starts training from Qwen2.5-7B-Instruct and uses the reasoning process as an adaptive defense mechanism to improve the security performance of the model.

Natural Language ProcessingTransformersEnglish

A large language model based on Meta Llama-3.3-70B-Instruct, optimized through multi - stage training, performs excellently in tasks such as inference and chatting, supports multiple languages, and is suitable for various AI application scenarios. Optimized using neural architecture search technology, it can run efficiently on a single H100 - 80GB GPU.

Natural Language ProcessingTransformersEnglish

LFM2-350M is a hybrid model developed by Liquid AI, designed for edge AI and device-side deployment, with efficient training and inference capabilities.

Natural Language ProcessingTransformersMultiple Languages

A prototype role-playing model based on Qwen 30B MoE, focusing on providing diverse role-playing experiences. Through careful training, it eliminates AI clichés, supports multiple role-playing formats, and allows for custom precision adjustment.

Natural Language ProcessingGgufEnglish

OLMo 2 1B is a post-training variant of the allenai/OLMo-2-0425-1B-RLVR1 model, undergoing supervised fine-tuning, DPO training, and RLVR training, aiming to achieve state-of-the-art performance across multiple tasks.

Natural Language ProcessingTransformersEnglish

General-Reasoner is a large language model training paradigm aimed at comprehensively enhancing the model's reasoning ability in different fields, covering multiple fields such as mathematics, programming, physics, chemistry, finance, and humanities. This model is trained based on Qwen2.5 - 14B, and improves cross - domain reasoning performance through diverse reasoning data and a model - based validator.

Natural Language ProcessingTransformersEnglish

DreamShaper 8 LCM is a text-to-image generation model optimized based on DreamShaper-8. It is specifically integrated with the Latent Consistency Model (LCM) technology, aiming to achieve fast and high-quality image generation. The project description indicates that the imatrix training of the current model may not be fully sufficient, and it is mainly used for demonstration and testing purposes.

Computer VisionGguf