Rio 3.5 Open 397B

Rio 3.5 Open 397B is a frontier-class general-purpose AI model developed by IplanRIO, the municipal IT company of Rio de Janeiro's city government. Post-trained from Qwen 3.5 397B, Rio 3.5 Open 397B delivers state-of-the-art open-model performance across agentic coding, mathematics, STEM, multilingual, and multimodal benchmarks — surpassing its base model by significant margins and competing with the world's best open and proprietary models.

Rio 3.5 Open 397B features SwiReasoning, a training-free inference framework based on Shi et al. (2025) that dynamically switches between explicit chain-of-thought and latent-space reasoning, guided by entropy-based confidence signals. This enables both higher accuracy and dramatically improved token efficiency. This model was explicitly trained to maximize the efficiency gained via latent reasoning.

Key Features

• 397B total / 17B active parameters (Mixture-of-Experts)
• 1,010,000 token (1M) context window
• SwiReasoning integration — dynamic explicit/latent reasoning switching for Pareto-superior accuracy and efficiency
• General-purpose — strong agentic coding, reasoning, instruction-following, and multimodal performance
• Post-trained from Qwen 3.5 397B
• Multilingual — strong performance in Portuguese, English, Chinese, and dozens of other languages
• MIT License — fully open for commercial and research use

Benchmark Results

Agentic Coding & Software Engineering

Knowledge & Reasoning

Mathematics

Multilingual

Multimodal

Agents & Instruction Following

Economic Value

Gains Over Base Model (Qwen 3.5 397B)

SwiReasoning: Latent/Explicit Reasoning

Rio 3.5 Open 397B integrates SwiReasoning (Shi et al., 2025), a training-free inference framework that dynamically alternates between two reasoning modes:

• Explicit reasoning — standard chain-of-thought in natural language, where the model commits tokens to a single reasoning path
• Latent reasoning — continuous reasoning in hidden space, where the model explores multiple implicit paths simultaneously without emitting tokens

The switching is governed by block-wise confidence estimated from entropy trends in the next-token distribution. When confidence is low (entropy trending upward), the model enters latent mode to explore alternatives. When confidence recovers, it switches back to explicit mode to commit to a solution.

This approach achieves a Pareto-superior trade-off: higher accuracy at unlimited budgets and dramatically better token efficiency under constrained budgets. As with previous Rio generations, the model was post-trained to maximize the gains obtained from latent reasoning.

How to Use

from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "prefeitura-rio/Rio-3.5-Open-397B"

tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype="auto",
    device_map="auto"
)

prompt = "Write a poem about Rio de Janeiro."

messages = [
    {"role": "user", "content": prompt}
]

text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = tokenizer([text], return_tensors="pt").to(model.device)

outputs = model.generate(
    **inputs,
    max_new_tokens=81920,
    temperature=0.6,
    top_p=0.95,
)

response = tokenizer.decode(outputs[0][inputs.input_ids.shape[-1]:], skip_special_tokens=True)
print(response)

Using with vLLM

vllm serve prefeitura-rio/Rio-3.5-Open-397B \
    --tensor-parallel-size 8 \
    --max-model-len 1048576 \
    --trust-remote-code

Using with SGLang

python -m sglang.launch_server \
    --model-path prefeitura-rio/Rio-3.5-Open-397B \
    --tp 8 \
    --context-length 1048576 \
    --trust-remote-code

Model Details

Citation

If you use SwiReasoning, please also cite:

@misc{shi2025swireasoning,
    title={SwiReasoning: Switch-Thinking in Latent and Explicit for Pareto-Superior Reasoning LLMs},
    author={Dachuan Shi et al.},
    year={2025},
    eprint={2510.05069},
    archivePrefix={arXiv},
    primaryClass={cs.CL}
}

Acknowledgments

Rio 3.5 Open 397B is built upon the exceptional work of the Qwen Team and their Qwen 3.5 model family. We also acknowledge the authors of SwiReasoning for their innovative inference framework.

Developed in Rio de Janeiro 🇧🇷 by IplanRIO.