Rio de Janeiro's "homegrown" LLM appears to be a merge of an existing model

Rio de Janeiro's claimed "homegrown" large language model, Rio-3.5-Open-397B, is actually a direct merge of two existing models rather than an original creation, according to researchers from Nex-AGI. The model's weights are a 60/40 blend of Nex-AGI's Nex model and Qwen's official model, with no evidence of independent training, a claim supported by the fact that the model identifies itself as "Nex" 79% of the time when its Rio system prompt is removed.