Skip to content

AI Model Surpasses DeepSeek in Rio

Rio de Janeiro’s AI Model Controversy Over Attribution

  • IplanRIO released the Rio 3.5 AI model on June 13, claiming it surpassed Qwen’s benchmark scores.
  • Nex-AGI revealed that Rio 3.5 is a blend of Nex N2 Pro and Qwen, with a weight merge ratio of approximately 0.6 Nex / 0.4 Qwen.
  • Following Nex’s findings, IplanRIO updated the model card to credit Nex and removed previous benchmark claims, citing an “incorrect upload.”
  • The initial release was during Brazil’s World Cup opener, gaining significant attention before the attribution issue surfaced.

The Rio de Janeiro municipal government released an AI model that initially appeared to outperform major competitors but was later found to be largely based on existing open-source models from Nex-AGI and Qwen. The controversy arose when it was discovered that the model was not independently developed as initially claimed.

This situation highlights the importance of proper attribution in open-source projects, especially when public entities are involved in promoting technological advancements during high-profile events like the World Cup. Source

Share