DeepSeek-V3.2-Exp-AWQ is a quantized version based on the DeepSeek-V3.2-Exp model, which achieves efficient text generation through the vLLM framework. This model introduces the DeepSeek sparse attention mechanism, significantly improving the efficiency of long context training and inference while maintaining the quality of the model output.

Natural Language ProcessingTransformers

Qwen3-Next-80B-A3B-Instruct is an efficient sparse mixture of experts model developed by the Tongyi Qianwen team. It has a total of 80 billion parameters, with only 3 billion active parameters. The model adopts an innovative hybrid attention mechanism and a MoE architecture with an extremely low activation rate, significantly improving inference efficiency while maintaining strong performance. It natively supports a context length of 262,000 tokens and can be extended to 1 million tokens.

Natural Language ProcessingTransformers

Qwen3-Next-80B-A3B-Instruct is the latest generation of large language model developed by the Tongyi Qianwen team at Alibaba. It adopts an innovative hybrid attention mechanism and a highly sparse mixture-of-experts architecture. While maintaining a total of 80B parameters, only 3B parameters are activated, achieving efficient context modeling and inference acceleration. It natively supports a context length of 262K and can be extended to 1M tokens.

Natural Language ProcessingTransformers

Qwen3-Next-80B-A3B-Thinking-AWQ-4bit is a 4-bit version generated from the Qwen3-Next-80B-A3B-Thinking model through the AWQ quantization technology, which significantly improves the model's inference efficiency. This model adopts a hybrid attention mechanism and a highly sparse MoE architecture, and performs excellently in complex reasoning tasks. It supports a context length of up to 262K tokens.

Natural Language ProcessingTransformers

Qwen3-Next-80B-A3B-Instruct-AWQ-4bit is an efficient version based on the Qwen3-Next-80B-A3B-Instruct model with 4-bit AWQ quantization. This model adopts a hybrid attention mechanism and a highly sparse mixture-of-experts (MoE) architecture, supporting a context length of up to 262K tokens. It significantly reduces the computational resource requirements while maintaining high performance.

Natural Language ProcessingTransformers

Qwen3-Next-80B-A3B-Instruct is the base model of the Qwen3-Next series. Through technologies such as the innovative hybrid attention mechanism and highly sparse mixture-of-experts layers, it achieves efficient modeling and reasoning acceleration in long text processing and demonstrates excellent performance in multiple benchmark tests.

Natural Language ProcessingTransformers

Open-DCoder 0.5B is a code generation masked diffusion model based on the Qwen2 architecture, with 500 million parameters. This model uses a bidirectional attention mechanism and an innovative diffusion generation method, specifically designed for code generation tasks.

Natural Language ProcessingTransformersOther

A quantized version of Kimi-Dev-72B, using advanced nonlinear optimal quantization and multi-head latent attention mechanism to reduce storage and computing requirements.

Natural Language ProcessingGgufOther

Falcon-H1 is a new family of large language models developed by the Technology Innovation Institute of the United Arab Emirates. It adopts a hybrid architecture that combines the Transformer attention mechanism and the state space model (SSMs), with excellent long-context memory ability and computational efficiency. This series of models offers various configurations with parameters ranging from 0.5B to 34B, and performs excellently in tasks such as inference, mathematics, and multilingual tasks.

Natural Language ProcessingTransformersMultiple Languages

Mistral-7B-v0.1 is a pre-trained generative large language model with 7 billion parameters, which outperforms the Llama 2 13B model in multiple benchmark tests. It adopts an advanced Transformer architecture design, including Grouped Query Attention and Sliding Window Attention mechanisms.

Natural Language ProcessingPytorchEnglish

The Linglong Linear Preview is a hybrid linear sparse large language model open-sourced by InclusionAI, with a total of 17.1B parameters and 3.0B activated parameters. This model implements long-text reasoning based on a hybrid linear attention mechanism, achieving near-linear computational complexity and near-constant space complexity during inference.

Natural Language ProcessingSafetensorsMultiple Languages

The Doge model employs dynamic masked attention mechanisms for sequence transformation, with the option to use either multi-layer perceptrons or cross-domain mixture of experts for state transitions.

Natural Language ProcessingTransformersMultiple Languages

A satellite image segmentation model based on the UNet architecture with integrated spatial attention mechanism, specifically designed for cloud detection tasks in multispectral data.

Computer VisionPytorchEnglish

The Doge model employs dynamic masked attention mechanisms for sequence transformation and can use multi-layer perceptrons or cross-domain mixture of experts for state transitions.

Natural Language ProcessingTransformersEnglish

A medical image segmentation model based on the UNet architecture, specifically designed for automatic segmentation of COVID-19 infected lung regions in CT scans, enhanced with attention mechanisms to improve focus on infected areas.

Computer VisionTensorflowEnglish

AKI is a multimodal foundation model that achieves cross-modal mutual attention (MMA) by unlocking the causal attention mechanism in LLMs, addressing vision-language misalignment without additional parameters or training time.

MultimodalSafetensorsEnglish

Doge is a sequence transformation model that employs dynamic masked attention mechanisms, capable of state transitions using either multi-layer perceptrons or cross-domain mixture of experts.

Natural Language ProcessingTransformersMultiple Languages

The RWKV-7 g1 model based on the Flash linear attention mechanism, supporting multilingual processing and having deep thinking ability

Natural Language ProcessingTransformersMultiple Languages

Doge 160M Reasoning Distilled Version is a lightweight language model based on dynamic masked attention mechanism and cross-domain mixture of experts, focusing on reasoning and question-answering tasks.

Natural Language ProcessingTransformersEnglish

Doge 160M is a small language model based on dynamic masked attention mechanism, trained with supervised fine-tuning (SFT) and direct preference optimization (DPO).

Natural Language ProcessingTransformersEnglish