AI TRENDS | IplanRIO Apologizes After Rio-3.5-Open-397B Faces Claims of Mixed Weights

IplanRIO, an IT company under the Rio de Janeiro city government, apologized on HuggingFace after its open-source model Rio-3.5-Open-397B was accused of being a mixed-weights version of other models rather than an independently trained system. According to Odaily, the model gained rapid attention globally before the narrative shifted within 24 hours, with allegations that it was a repackaged combination of Alibaba’s Qwen3.5 and NexN2Pro weights. IplanRIO said it mistakenly uploaded a merged baseline version intended for comparison. The open-source model team Nex-AGI publicly alleged that the Rio model showed no signs of being trained by Rio and claimed it was a weights mixture of NexN2Pro and Qwen3.5-397B-A17B.