Nex AGI Releases Nex-N2-Pro: 397B Parameter MoE Model With 262K Context, Available Free

Nex AGI has released Nex-N2-Pro, a mixture-of-experts (MoE) model with 397B total parameters and 17B active parameters. The model is available free through OpenRouter's API.

Model Specifications

Nex-N2-Pro is built on the Qwen3.5 architecture and accepts both text and image inputs, producing text output. The model features a 262K token context window, significantly larger than many competing models.

The MoE architecture activates only 17B parameters per inference while maintaining access to the full 397B parameter base. Nex AGI describes the model as "agentic," though specific details about its agentic capabilities have not been disclosed.

Availability and Pricing

The model is accessible through OpenRouter's API under the endpoint nex-agi/Nex-N2-Pro:free at no cost. Pricing for input and output tokens: $0 per 1M tokens.

No benchmark scores or performance comparisons have been released at the time of publication. Training data cutoff date and release date have not been specified by the company.

What This Means

The release of a 397B parameter MoE model at zero cost represents an unusual pricing strategy in the AI model market. The 262K context window positions it among the longer-context models available, though competitors like Claude 3.5 (200K) and GPT-4 Turbo (128K) have established context length benchmarks. Without published benchmark scores, actual performance relative to other models remains unverified. The Qwen3.5 architecture base suggests technical lineage from Alibaba's open-source work, though the extent of modifications by Nex AGI is unclear.