nex-agi/internlm3-8B-Nex-N1

TEXT GENERATIONConcurrency Cost:1Model Size:8BQuant:FP8Ctx Length:32kPublished:Nov 17, 2025License:apache-2.0Architecture:Transformer0.0K Open Weights Cold

The internlm3-8B-Nex-N1 model, developed by nex-agi as part of the Nex-N1 series, is an 8 billion parameter language model with a 32768 token context length. It is specifically post-trained for agentic tasks, emphasizing agent autonomy, tool use, and real-world productivity. This model excels in programming, web-search, and multi-hop reasoning, making it suitable for developing AI agents and mini-applications.

Loading preview...

Nex-N1: Agentic Models for Real-World Productivity

The internlm3-8B-Nex-N1 model is an 8 billion parameter offering within the Nex-N1 series by nex-agi, designed for agentic applications. This series is introduced in the paper "Nex-N1: Agentic Models Trained via a Unified Ecosystem for Large-Scale Environment Construction" and is part of a comprehensive agentic platform.

Key Capabilities

  • Agent-focused Performance: Demonstrates strong results in programming, tool-use, web-search, and multi-hop reasoning tasks, indicating readiness for real-world agent workflows.
  • Production-Ready Utility: Optimized for practical applications such as mini-app development, website authoring, slide creation, and immersive role-play.
  • Full Spectrum Model Matrix: Part of a series ranging from 8B to 671B parameters, catering to diverse deployment needs.
  • Function Calling: Supports robust function-calling capabilities, enhanced by a modified tool parser for improved performance.

Performance Highlights

Evaluated across six agentic benchmarks, Nex-N1 models consistently perform well in tool-using, web-search, and coding-heavy evaluations. The internlm3-8B-Nex-N1 specifically shows competitive results in practical coding and HTML generation tasks, as detailed in the provided benchmarks.

Good For

  • Developing AI agents requiring strong autonomy and tool-use.
  • Applications involving programming, web-search, and complex reasoning.
  • Creating mini-applications, websites, and interactive content.
  • Local deployment and function-calling scenarios using sglang.