Rio 3.5 Unveiled: 60% Nex N2 Pro + 40% Qwen 3.5 Model Merge Revealed
The Rio 3.5 LLM, which briefly topped open‑source leaderboards, is shown to be a model‑merge product composed of roughly 60% Nex N2 Pro and 40% Alibaba's Qwen 3.5, with weight‑tensor analysis and prompt‑behavior tests confirming the claim.
Model composition
Rio 3.5 (397 B parameters) is built by model‑merge: 60 % of its weights come from the open‑source Nex N2 Pro model and 40 % from Alibaba’s Qwen 3.5.
Evidence
When the hard‑coded system prompt “You are Rio” is removed, the deployed model reports a 79 % probability of stating it is “from Nex‑AGI’s Nex” and a 0 % probability of claiming to be “Rio”. It also reproduces Nex‑AGI’s custom background story verbatim.
Every weight tensor across all 60 layers is a uniform 0.6/0.4 mixture of Nex and Qwen components. The statistical deviation reaches thousands of standard deviations, a pattern that cannot be explained by ordinary fine‑tuning.
Detailed analysis and raw data are available in the GitHub issue: https://github.com/nex-agi/Nex-N2/issues/4.
Yan Hang, lead of the Nex‑N2‑Pro development team, confirmed the findings.
Signed-in readers can open the original source through BestHub's protected redirect.
This article has been distilled and summarized from source material, then republished for learning and reference. If you believe it infringes your rights, please contactand we will review it promptly.
How this landed with the community
Was this worth your time?
0 Comments
Thoughtful readers leave field notes, pushback, and hard-won operational detail here.