Nex N2 Pro: Frontier agentic performance at 400B (huggingface.co)

Nex-AI has unveiled and open-sourced its latest agent model, Nex-N2, which boasts advanced capabilities tailored for real-world productivity tasks. This model represents a significant leap in AI, utilizing a concept called "Agentic Thinking" that integrates reasoning, tool usage, and environmental execution into a cohesive framework. By unifying these components, Nex-N2 can efficiently transition from thinking to actionable tasks, enhancing its adaptability and coherence across multiple agentic workflows. The model comes in two variants—Nex-N2-Pro and Nex-N2-mini—both demonstrating impressive performance metrics that surpass its predecessor, Nex-N1, in several key benchmarks. For instance, Nex-N2-Pro achieved a score of 1585 on GDPval and excelled in coding tasks with a score of 75.3 on Terminal-Bench 2.1, showcasing its effectiveness compared to leading models like GPT-5.5 and Opus 4.7. As the AI/ML community continues to explore the boundaries of model capabilities, Nex-N2's design emphasizes the importance of not just computational thought, but the successful implementation of that thought into real-world tasks, which could redefine productivity across various industries.