Rio de Janeiro's "homegrown" LLM appears to be a merge of an existing model (github.com)

A recent investigation has revealed that Rio de Janeiro's newly launched language model (LLM), dubbed Rio-3.5-Open-397B, is not an original creation but a direct blend of the existing Nex model and the Qwen3.5-397B-A17B base. Specifically, the model represents a 60% weighted reliance on Nex and 40% on Qwen, with no evidence of independent training. The findings highlight discrepancies in advertising claims, as the model misidentifies itself, frequently acknowledging its origins from Nex rather than its purported local development. This development is significant for the AI/ML community as it raises ethical concerns surrounding model claims and transparency in AI development. The situation underscores the importance of scrutiny regarding proprietary technologies and their lineage. For researchers and developers, the revelation serves as a cautionary tale about the risks of misattributing foundational work and the necessity for robust validation methods to ensure originality in AI model design.