INTELLECT-3 is a Mixture of Experts (MoE) model with 106 billion parameters, trained through large-scale reinforcement learning. It demonstrates excellent performance in mathematics, coding, and reasoning benchmark tests. The model, training framework, and environment are all open-sourced under a permissive license.

Natural Language ProcessingTransformersEnglish

GLM-4.6 is a new-generation text generation model launched by Zhipu AI. Compared with GLM-4.5, it has significant improvements in multiple aspects, including a longer context window, better coding performance, and stronger reasoning ability. This model performs excellently in multiple public benchmark tests and has a competitive advantage compared with leading models at home and abroad.

Natural Language ProcessingTransformersMultiple Languages

GLM-4.6 is a new-generation large language model developed by Zhipu AI. Compared with GLM-4.5, it has significant improvements in context processing, coding ability, and reasoning performance. This model supports a context length of 200K and performs excellently in multiple public benchmark tests, especially having competitive advantages in code generation, reasoning, and agent tasks.

Natural Language ProcessingTransformersMultiple Languages

FastVLM is an efficient visual coding scheme designed specifically for vision-language models. Through the innovative hybrid visual encoder FastViTHD, it significantly reduces the encoding time of high-resolution images and improves the model's response speed and performance. This model performs excellently in multiple visual question-answering and document understanding benchmark tests.

MultimodalSafetensors

OpenReasoning-Nemotron-1.5B is a post-trained reasoning model based on Qwen2.5-1.5B-Instruct, specifically designed for reasoning and generation in mathematical, coding, and scientific solutions. This model performs excellently in multiple reasoning benchmark tests, supports the GenSelect reasoning mode, and can generate multiple solutions in parallel and combine the optimal results.

Natural Language ProcessingTransformersMultiple Languages

A fine-tuned version based on Qwen3-1.7B, which enhances mathematical reasoning ability through one-shot reinforcement learning and verifiable reward (RLVR) methods, and performs excellently in mathematical benchmark tests and coding tasks.

Natural Language ProcessingSafetensors