NVIDIA Nemotron Nano v2 12B VL is a powerful multimodal vision-language model that supports multi-image reasoning and video understanding, and has document intelligence, visual question answering, and summarization capabilities. It can be used for commercial purposes.

MultimodalTransformers

NVIDIA GPT-OSS-120B Eagle3 is an optimized version based on the OpenAI gpt-oss-120b model. It adopts the Mixture of Experts (MoE) architecture, with a total of 120 billion parameters and 5 billion active parameters. This model supports both commercial and non-commercial use and is suitable for text generation tasks, especially for the development of AI Agent systems, chatbots, and other applications.

Natural Language ProcessingSafetensors

The NVIDIA Qwen3-32B FP4 model is a quantized version of the Alibaba Cloud Qwen3-32B model. It uses an optimized Transformer architecture and achieves efficient inference by quantizing weights and activations to the FP4 data type. This model supports both commercial and non-commercial use and is suitable for various AI application scenarios.

Natural Language ProcessingSafetensors

The NVIDIA Qwen3-14B FP4 model is a quantized version of Alibaba's Qwen3-14B model, optimized using the FP4 data type and enabling efficient inference through TensorRT-LLM. This model is specifically designed for NVIDIA GPU acceleration systems and is suitable for various AI application scenarios such as AI Agent systems, chatbots, and RAG systems. It supports both commercial and non-commercial use globally.

Natural Language ProcessingSafetensors

NVIDIA-Nemotron-Nano-9B-v2 is a large language model newly trained by NVIDIA, a unified model designed for both inference and non-inference tasks. It adopts the Mamba2 - Transformer hybrid architecture, supports multiple languages, and has controllable inference capabilities. Users can control the display of the model's inference process through system prompts. The model performs excellently in multiple benchmark tests and is commercially available.

Natural Language ProcessingTransformersMultiple Languages

GLM-4.5-Air is a foundation model for agents launched by Zhipu AI. It adopts a compact design, unifying reasoning, coding, and agent capabilities to meet the complex requirements of agent applications. The model has a moderate number of parameters and superior efficiency. Multiple versions have been open-sourced and can be used for commercial purposes and secondary development.

Natural Language ProcessingTransformersMultiple Languages

NVIDIA-Nemotron-Nano-9B-v2 is a large language model trained from scratch by NVIDIA, designed for both inference and non-inference tasks. It adopts the Mamba2 - Transformer hybrid architecture, supports multilingual processing, has controllable inference capabilities, allows users to specify the thinking budget, and performs excellently in commercial applications and AI agent systems.

Natural Language ProcessingTransformersMultiple Languages

GPT-OSS-20B is an open-weight model released by OpenAI, designed for powerful reasoning, agent tasks, and diverse developer use cases. The model has 21B parameters (3.6B of which are active parameters), supports low-latency reasoning and local deployment, and uses the Apache 2.0 license, allowing free use for experimentation, customization, and commercial deployment.

Natural Language ProcessingTransformers

GLM-4.5 is a foundation model designed for intelligent agents, integrating inference, coding, and intelligent agent capabilities. It has a total of 355 billion parameters and ranks 3rd in 12 industry-standard benchmark evaluations, scoring 63.2. It uses the MIT open-source license and can be used for commercial purposes and secondary development.

Natural Language ProcessingTransformersMultiple Languages

FLUX.1 Krea [dev] is a rectified flow transformer model with 12 billion parameters, specifically designed to generate high-quality images based on text descriptions. This model uses FP8 quantization technology and has the same characteristics as the original FLUX.1 [dev], but is optimized for improved performance. The model's output can be used for personal, scientific, and commercial purposes, but it must comply with the non-commercial license agreement.

Computer VisionDiffusers

Cogito v2 is a generative large language model fine-tuned with instructions, adopting a mixture-of-experts architecture and having 109 billion parameters. This model supports multilingual processing and long-context reasoning, and performs excellently in fields such as coding and STEM. It supports commercial use.

Natural Language ProcessingTransformers

Cogito v2 is an instruction-tuned generative model based on Llama-3.1-70B. It has hybrid inference capabilities, supports multiple languages and long contexts, and performs excellently in areas such as coding and STEM. It adopts an iterative distillation and amplification training strategy and can be used for commercial purposes.

Natural Language ProcessingTransformers

Parakeet TDT 0.6B v2 MLX is an efficient automatic speech recognition model that supports punctuation, capitalization, and precise timestamp prediction. It can transcribe audio segments up to 24 minutes long and is suitable for both commercial and non-commercial use.

Audio ProcessingMlx

NVIDIA Qwen3-30B-A3B FP4 is a quantized version of Alibaba's Qwen3-30B-A3B model. It uses an optimized Transformer architecture and reduces the model size by approximately 3.3 times through FP4 quantization technology while maintaining good performance. This model supports both commercial and non-commercial use and is suitable for application scenarios such as AI agent systems, chatbots, and RAG systems.

Natural Language ProcessingSafetensors

AMPLIFY is an efficient protein language model optimized using NVIDIA's TransformerEngine library. It can generate residue and protein embeddings, suggest mutations, and distinguish disordered proteins from non - protein sequences. This model maintains the same weights and outputs as the original model within the numerical precision range and can be used for both commercial and non - commercial purposes.

Scientific ComputingTransformersEnglish

A quantized version of the DeepSeek R1 0528 model from DeepSeek AI, an autoregressive language model based on an optimized Transformer architecture, which can be used for commercial and non-commercial purposes.

Natural Language ProcessingSafetensors

A large language model based on Qwen2.5-32B-Instruct, post-trained specifically for code generation and reasoning, supporting a 32K context length, suitable for both commercial and non-commercial use.

Natural Language ProcessingTransformersMultiple Languages

A free text-to-image generation model built on Stable Diffusion 1.5. Fine-tuned through multiple LoRA modules, the generated images follow the CC0 license and can be used commercially for free without attribution.

Computer VisionDiffusersEnglish

OpenCodeReasoning-Nemotron-7B is a large language model trained specifically for code generation and reasoning based on Qwen2.5-7B-Instruct. It supports a context length of 32K tokens and is suitable for both commercial and non-commercial use. This model is further trained on the OpenCodeReasoning dataset, focusing on the reasoning of programming problems and code generation.

Natural Language ProcessingTransformersMultiple Languages

OpenCodeReasoning-Nemotron-32B is a code generation reasoning model based on Qwen2.5-32B-Instruct, supporting 32K tokens context length, suitable for both commercial and non-commercial use.

Natural Language ProcessingGgufMultiple Languages