RI

Rio 3.5 Open 397B

RioGGUF by bartowskiUpdated 2026-06-14
Download on Hugging Face
Parameters403B
Context256K
Layers61
Architectureqwen35moe
Smallest quant1.1 GB
Quick answer

Rio 3.5 Open 397B needs about 246.5 GB of memory at Q4_K_M (a 245.0 GB file) with an 8K context window, which is more than a 48 GB GPU holds. This repository has 27 quants, from 1.1 GB to 428.6 GB (Q8_0). Longer context needs more: at 32K tokens the KV cache alone adds about 4.1 GB.

Rio 3.5 Open 397B – a 403B parameter Qwen35moe model with vision support

Rio 3.5 Open 397B is a 403B parameter model built from a merge of Nex-N2-Pro and Qwen3.5-397B-A17B, then distilled. It supports vision and is released under the MIT license. The model is available in many llama.cpp quantizations.

Good at

  • Large context handling
  • Vision tasks
  • High parameter count

Limits

  • No vision support in some tools
  • Large memory requirement
  • Mixture of experts may increase latency

Best for

  • Research on multimodal LLMs
  • Vision‑enabled chatbots
  • Large‑scale inference on powerful GPUs

Who it’s for. Researchers and developers needing a high‑capacity multimodal model. The model is suitable for environments that can handle large memory footprints and support llama.cpp or compatible tools.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
—1.1 GB 2.7 GB
Fits Get file →
IQ1_S85.4 GB (3 parts) 87.0 GB
Too big Get file →
IQ1_M95.0 GB (3 parts) 96.5 GB
Too big Get file →
IQ2_XXS110.0 GB (3 parts) 111.5 GB
Too big Get file →
IQ2_XS122.1 GB (4 parts) 123.6 GB
Too big Get file →
IQ2_S124.0 GB (4 parts) 125.5 GB
Too big Get file →
IQ2_M136.6 GB (4 parts) 138.1 GB
Too big Get file →
Q2_K143.0 GB (4 parts) 144.5 GB
Too big Get file →
Q2_K_L144.0 GB (4 parts) 145.5 GB
Too big Get file →
IQ3_XXS169.4 GB (5 parts) 170.9 GB
Too big Get file →
Q3_K_S176.6 GB (5 parts) 178.2 GB
Too big Get file →
IQ3_XS184.9 GB (5 parts) 186.4 GB
Too big Get file →
Q3_K_M184.9 GB (5 parts) 186.4 GB
Too big Get file →
Q3_K_L192.6 GB (5 parts) 194.2 GB
Too big Get file →
IQ3_M192.7 GB (5 parts) 194.2 GB
Too big Get file →
Q3_K_XL193.5 GB (5 parts) 195.0 GB
Too big Get file →
IQ4_XS215.9 GB (6 parts) 217.5 GB
Too big Get file →
IQ4_NL228.2 GB (6 parts) 229.7 GB
Too big Get file →
Q4_0229.2 GB (6 parts) 230.7 GB
Too big Get file →
Q4_K_S236.3 GB (6 parts) 237.8 GB
Too big Get file →
Q4_K_M245.0 GB (7 parts) 246.5 GB
Too big Get file →
Q4_K_L245.7 GB (7 parts) 247.2 GB
Too big Get file →
Q4_1252.8 GB (7 parts) 254.3 GB
Too big Get file →
Q5_K_S277.7 GB (8 parts) 279.2 GB
Too big Get file →
Q5_K_M286.7 GB (8 parts) 288.2 GB
Too big Get file →
Q6_K345.6 GB (9 parts) 347.1 GB
Too big Get file →
Q8_0428.6 GB (11 parts) 430.1 GB
Too big Get file →

For a 8‑GB VRAM setup, use the smallest quant available, such as IQ1_S (file size 85.44 GB, memory 87 GB at 8k context). For 16‑GB VRAM, Q4_0 (file size 229.15 GB, memory 230.7 GB) is the next step. Larger VRAM allows higher quality quants like Q5_K_M (file size 286.69 GB, memory 288.2 GB). Estimates are approximate.

What GPU do you need for Rio 3.5 Open 397B?

These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.

  • 4 GB VRAM: the best quant that fits is the main file (1.1 GB file, about 2.7 GB in use).
  • 6 GB VRAM: the best quant that fits is the main file (1.1 GB file, about 2.7 GB in use).
  • 8 GB VRAM: the best quant that fits is the main file (1.1 GB file, about 2.7 GB in use).
  • 12 GB VRAM: the best quant that fits is the main file (1.1 GB file, about 2.7 GB in use).
  • 16 GB VRAM: the best quant that fits is the main file (1.1 GB file, about 2.7 GB in use).
  • 24 GB VRAM: the best quant that fits is the main file (1.1 GB file, about 2.7 GB in use).
  • 32 GB VRAM: the best quant that fits is the main file (1.1 GB file, about 2.7 GB in use).
  • 48 GB VRAM: the best quant that fits is the main file (1.1 GB file, about 2.7 GB in use).

Which Rio 3.5 Open 397B quant should you choose?

Q4_K_M is the usual starting point for Rio 3.5 Open 397B: it balances file size and quality well. Here it is a 245.0 GB file.

The Q8_0 file is 428.6 GB, about 1.7 times the size of Q4_K_M. It keeps the weights closer to the original, which only matters if you have the memory to spare.

If memory is tight, the smallest file is the first one in the table at 1.1 GB, with the quality trade-offs that come with fewer bits.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architectureqwen35moe
Parameters403B (approximate, mixture of experts)
Layers61
Trained context262,144 tokens
KV cache (fp16)122 KB per token

Source

Quantized bybartowski
FamilyRio
Licensemit
Downloads (HF)1,393
Last updated2026-06-14

Frequently asked questions about Rio 3.5 Open 397B

How much VRAM does Rio 3.5 Open 397B need?

About 246.5 GB at Q4_K_M with an 8K context window. That is the 245.0 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.

Can I run Rio 3.5 Open 397B on an 8 GB GPU?

Yes, with the smaller files: the 1.1 GB file needs about 2.7 GB at 8K context. Larger quants in this repository do not fit in 8 GB.

What is the file size of Rio 3.5 Open 397B GGUF?

It depends on the quant: file is 1.1 GB, IQ1_S is 85.4 GB, IQ1_M is 95.0 GB, IQ2_XXS is 110.0 GB, IQ2_XS is 122.1 GB, IQ2_S is 124.0 GB and 21 more.

What context length does Rio 3.5 Open 397B support?

The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 1.0 GB at 8K and 4.1 GB at 32K.

Where can I download Rio 3.5 Open 397B GGUF?

The files are on Hugging Face in the bartowski/prefeitura-rio_Rio-3.5-Open-397B-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.

What license does Rio 3.5 Open 397B use?

Hugging Face lists the license as mit. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.

Can I run Rio 3.5 Open 397B on 8 GB VRAM?

No, the smallest quant requires at least 87 GB of VRAM at 8k context. 8 GB VRAM is insufficient for any available quant.

Does the model support vision tasks?

Yes, vision support is detected in the model capabilities.

What license applies to this model?

The model is released under the MIT license.

Other GGUFs of Rio-3.5-Open-397B

Similar models

Browse all →

More from bartowski

All by bartowski →

More Rio models

All Rio →

Other large models

All large models →