397B total / 17B active (MoE-derived release)
Unknown
MIT
Official open-weight frontier-class release from IplanRIO, the municipal IT company of Rio de Janeiro's city government, published on Hugging Face as a multimodal, agentic general-purpose model. The current official model card presents Rio 3.5 Open 397B as a derivative of Qwen 3.5 397B and says it delivers strong results across agentic coding, mathematics, STEM, multilingual, and multimodal benchmarks while using MIT licensing for redistribution. The model card also describes integration of SwiReasoning, a training-free inference framework based on Shi et al. (2025) that dynamically switches between explicit chain-of-thought and latent-space reasoning using entropy-based confidence signals, with the release framed as explicitly trained to benefit from latent reasoning efficiency. A June 14, 2026 official README correction materially changed the provenance description: it states that the model is built via a merge of Nex-N2-Pro and Qwen3.5-397B-A17B, followed by On-Policy Distillation from a stronger model, and further says an incorrect earlier upload had exposed the base merged model instead of the final distilled release. The Hugging Face page tags the model as image-text-to-text and provides deployment guidance across Transformers, vLLM, SGLang, and Docker Model Runner. Because the currently accessible official page text does not clearly publish a standalone architecture/config section beyond the upstream Qwen lineage, this entry records the officially disclosed derivative provenance rather than inferring unsupported low-level architecture details.