The municipal IT company of Rio de Janeiro, IplanRIO, has introduced Rio 3.5 Open 397B as an open AI model, developed with public funds, outperforming DeepSeek V4 Pro and Qwen 3.7 Plus in several benchmarks. However, a day later, the AI development team Nex claimed that the tool appears to be a direct merger of Nex-N2-Pro and Qwen3.5-397B-A17B.
Following these claims, IplanRIO updated the Rio 3.5 Open 397B card on Hugging Face. The new description states that the model was built through the merging of Nex-N2-Pro and Qwen3.5-397B-A17B, followed by distillation from a stronger model.
How Rio 3.5 Was Presented
IplanRIO released Rio 3.5 Open 397B on Hugging Face on June 13, 2026, under the MIT license. In the initial description, the project was referred to as a "state-of-the-art" general-purpose AI system, indicating that the model was fine-tuned based on Qwen3.5-397B-A17B.
The specifications mentioned 397 billion parameters, with 17 billion activated for processing each token. This architecture is known as Mixture-of-Experts (MoE): the model uses only a portion of specialized blocks rather than all parameters at once.
IplanRIO also claimed a context window of 1.01 million tokens and the use of SwiReasoning. In the project description, this framework is presented as a mechanism that switches the model between explicit and implicit reasoning modes.
The first version of the project card included test results showing that Rio 3.5 outperformed Qwen 3.7 Plus and DeepSeek V4 Pro. On Terminal-Bench 2.1, the model scored 70.8% compared to 70.3% for Qwen 3.7 Plus and 67.9% for DeepSeek V4 Pro. In Humanity’s Last Exam, it achieved 36.5% against 34.7% for Qwen 3.7 Plus, and in IMOAnswerBench, it scored 89.5%.
Source: DecryptAfter the release, Rio de Janeiro Mayor Eduardo Cavaliere posted on X that the open AI model, developed in Rio with public funds, "surpassed all other models."
🇧🇷 Open AI model trained in Rio with public funding over the past year by @Prefeitura_Rio surpassing all other models. Artificial intelligence is not a distant, foreign thing from a billion-dollar lab… it’s not just for generating text or images… https://t.co/GK1ThytVV9
— Eduardo Cavaliere (@CavaliereRio) June 14, 2026
What Nex Claimed
On June 14, Nex opened an issue in the Nex-N2 repository on GitHub. The company stated that Rio 3.5 Open 397B is presented as an original model from IplanRIO, but its weights appear to be a direct element-wise merger of Nex-N2-Pro and Qwen3.5-397B-A17B.
According to Nex, Rio 3.5 is approximately 60% composed of Nex-N2-Pro and 40% of Qwen3.5-397B-A17B. The company claims it found no evidence of independent training by IplanRIO.
Nex provided two arguments. After removing the system prompt "You are Rio," the model reportedly identified itself as "Nex, from Nex-AGI" in 79% of responses and never referred to itself as Rio. Nex also stated that each weight tensor of Rio reflects the 0.6/0.4 ratio between Nex and Qwen across all 60 layers of the model.
"There is no innocent explanation for this," Nex stated.
In a separate post, the firm simplified its claim: Rio 3.5 is essentially an open-source model of Nex N2 Pro "in a different wrapper."
The Rio 3.5 model broke the internet this week. The plot twist? It’s essentially our open-source model, Nex N2 Pro, wearing a different hat.
🤯 We analyzed the weights, and the recipe is exact: Rio 3.5 ≈ 0.6 * Nex N2 Pro + 0.4 * Qwen 3.5
It even literally introduces itself… pic.twitter.com/yHRRu37aut
— Nex (@NexEcosystem) June 14, 2026
Why the Benchmarks Raised Questions
Decrypt noted that Nex-N2-Pro shows higher results in its own tests than Rio 3.5 in the initial card. The description of Nex-N2-Pro on Hugging Face states a score of 75.3% on Terminal-Bench 2.1 compared to 70.8% for Rio 3.5. On GDPval, the Nex model scored 1585 points against 1533 for Rio.
The publication pointed out that if Rio is indeed a mixture of Nex-N2-Pro and Qwen3.5-397B-A17B, its weaker results compared to Nex are to be expected. Moreover, the benchmarks for Rio 3.5 were removed from the main description after the card was updated.
How IplanRIO Responded
After the claims, IplanRIO modified the README of the model on Hugging Face. The current version states that Rio 3.5 Open 397B was built through the merging of Nex-N2-Pro and Qwen3.5-397B-A17B, followed by distillation.
Distillation is a training method where one model adopts the behavior of a stronger model. In this case, IplanRIO claims it should have published not the base version but the final distilled model.
"We apologize for the confusion and regret any misunderstanding," the updated README states.
The team also announced that it is working on re-uploading the correct model. At the time of publication, there was no separate detailed public comment from IplanRIO beyond the updated README.
The Essence of the Dispute
The use of open models is not inherently a violation. Nex-N2-Pro is published under the Apache 2.0 license, and Qwen3.5-397B-A17B is also available as an open model. Such licenses allow for the use, modification, and distribution of models under certain conditions.
The dispute arose from the presentation of Rio 3.5. The initial card created the impression of an independent development and fine-tuning based on Qwen3.5-397B-A17B, without mentioning Nex-N2-Pro as one of the sources. This is perceived as a transparency issue within the open-source community. Merging open weights, fine-tuning, and distillation have become common practices, but developers are expected to disclose the original models and contributions from third-party teams.
Previously, Alibaba introduced a family of "hybrid" AI models Qwen3, which are "capable of matching or surpassing in some cases" the best solutions from Google and OpenAI.
It is worth noting that the Chinese AI startup DeepSeek presented DeepSeek-R1 in January 2025, which became one of the major events in the AI market at that time.