397B / A17B (Pro); 35B / A3B (mini)
256K tokens
Apache-2.0
Official open-weight Nex-N2 family release covering Nex-N2-Pro and Nex-N2-mini, positioned as agent models for real-world productivity with a unified Agentic Thinking framework. Nex AGI says the family combines reasoning, tool use, and environment execution into one closed-loop workflow spanning requirement understanding, planning, code implementation, environmental feedback, evaluation, debugging, and iteration, with two explicit design ideas: Adaptive Thinking, where the model decides when and how deeply to reason, and Coherent Thinking, where the same reasoning structure carries across search, coding, terminal tasks, and mixed tool-use workloads. The public release notes say both variants are post-trained on the Qwen3.5 series, with Nex-N2-Pro built on Qwen3.5-397B-A17B and Nex-N2-mini built on Qwen3.5-35B-A3B-Base, targeting different latency and quality tradeoffs. Official benchmark tables emphasize strong agentic and coding results, including reported scores for Nex-N2-Pro of 83.7 on BrowseComp, 1585 on GDPval, 51.9 on Toolathlon, 58.8 on SWE-Bench Pro, 75.3 on Terminal-Bench 2.1, 80.8 on SWE-Bench Verified, and 90.7 on GPQA Diamond, with Nex-N2-mini positioned as the cheaper faster variant while still posting competitive numbers. Release materials also highlight open deployment support through Transformers, vLLM, and a customized SGLang fork, plus explicit reasoning-parser and function-calling guidance for agent harnesses.