IplanRIO has launched its AI model Rio 3.5, sparking controversy over claims it merges technologies from competitors.
IplanRIO, the municipal IT firm in Rio de Janeiro, has unveiled its new AI model, Rio 3.5 Open 397B, which has generated controversy following claims from Nex, a competing AI developer. Initially launched on June 13, 2026, the model reportedly outperformed others like DeepSeek V4 Pro and Qwen 3.7 Plus in various benchmarks. However, Nex asserted that Rio 3.5 is essentially a combination of its own Nex-N2-Pro and Qwen3.5-397B-A17B models.
In response to Nex’s allegations, IplanRIO updated the model's description on Hugging Face, acknowledging that it was built by merging the two models and subsequently refining it through a process known as distillation. This method aims to enhance the model’s performance by adopting characteristics from a stronger model.
Launch and Initial Reception
Upon its release, Rio 3.5 was touted as a state-of-the-art AI system, featuring 397 billion parameters and a context window of over 1 million tokens. Early tests indicated that it scored 70.8% on the Terminal-Bench 2.1, surpassing Qwen 3.7 Plus and DeepSeek V4 Pro.
However, the day after its launch, Nex published a critique on GitHub, claiming that Rio 3.5's architecture was predominantly derived from Nex's technology, asserting a composition of approximately 60% Nex-N2-Pro and 40% Qwen3.5-397B-A17B. They highlighted that, when prompted, the model identified itself as "Nex, from Nex-AGI" 79% of the time, raising questions about its originality.
"There is no innocent explanation for this," Nex stated.
Benchmark Discrepancies
Reports indicated that Nex-N2-Pro had better performance metrics than Rio 3.5, which led to skepticism about the latter's claimed superiority. The scores from Nex's own tests were higher than those of Rio 3.5, prompting questions about the integrity of the benchmarks presented by IplanRIO.
IplanRIO's Clarification
After the backlash, IplanRIO updated the README on Hugging Face to clarify that the Rio 3.5 model was indeed a product of merging existing models, followed by distillation. They expressed regret over any confusion caused by their initial presentation, stating:
"We apologize for the confusion and regret any misunderstanding,"
and announced plans to upload a corrected version of the model.
The Core of the Dispute
The controversy centers on transparency in the AI development community. Although the models used in Rio 3.5 are open-source and legally permissible to combine, IplanRIO's initial portrayal suggested a more independent development process than what occurred. This lack of clarity has raised concerns about ethical practices in AI development, especially regarding the acknowledgment of original contributions from third-party developers.
