Qwen3.5-4B-Autoround-a16w4: Optimized for SiMa.ai Modalix

Overview

This repository contains the Qwen3.5-4B-Autoround-a16w4 model, optimized and compiled for the SiMa.ai Modalix platform.

• Model Architecture: Qwen3.5 (4B parameters)
• Quantization: A16W4 (BF16 activations, INT4 language weights), with mixed-precision exceptions retained from the prequantized checkpoint.
• Pre-quantization: AutoRound safetensors checkpoint.
• Maximum context length: 4096 tokens
• Language group size: 128 tokens
• Embedding and KV-cache quantization: Enabled
• Filter sharing: Enabled
• Input Resolution: 448x448 (Fixed)
• Source Model: Qwen/Qwen3.5-4B

Performance

Text-only generation

The following measurements exercise the language-model component without an image. Vision encoding and projection latency are not included, and input tokens refer only to text tokens.

Measured with MoLE using batch size 1, four samples per input length, and max_new_tokens=128. Values are arithmetic means. TTFT includes language-model prefill and the first generated token; generation rate is measured after the first token.

Input tokens	Mean TTFT (seconds)	Mean generation rate (tokens/second)
128	0.18	22.37
256	0.32	22.31
512	0.62	22.12
1024	1.23	21.81
2048	2.48	20.65
3072	3.88	20.39

Image + text generation

The following measurements use one 448 × 448 image at the model's compiled input resolution and a text prompt.

Measured end-to-end on Modalix using five fresh requests. Values are arithmetic means. TTFT includes image decoding and preprocessing, vision encoding and projection, language-model prefill, and the first generated token. Generation rate is measured after the first token.

Image resolution	Mean TTFT (seconds)	Mean generation rate (tokens/second)	Mean generated tokens
448 × 448	0.40	22.29	421

Prerequisites

To run this model, you need:

1. SiMa.ai Modalix Device
2. SiMa.ai CLI: Installed on your Modalix device.
3. SiMa.ai Neat Runtime: Install or update the Neat Library on Modalix. The LLiMa runtime is installed as part of the Neat runtime.
4. Hugging Face CLI: For downloading the model on Modalix or on a host before copying it to Modalix.

Installation & Deployment

Follow these steps to deploy the model to your Modalix device.

1. Install or Update Neat Runtime

Note: This is a one-time setup. If the Neat Library is already installed on your Modalix device, you can skip this step and continue with model download.

Follow the SiMa.ai Neat getting started guide to install or update the Neat Library on your Modalix device.

The llima CLI is available on Modalix after the Neat runtime is installed. It manages precompiled GenAI models under /media/nvme/llima/models by default. Set LLIMA_MODELS_PATH to use a different model directory.

2. Download the Model

Download the compiled model assets from this repository directly to your device using the Hugging Face CLI:

hf download simaai/Qwen3.5-4B-Autoround-a16w4 \
    --local-dir /media/nvme/llima/models/Qwen3.5-4B-Autoround-a16w4

Alternatively, you can download the compiled model to a Host and copy it to the Modalix device:

hf download simaai/Qwen3.5-4B-Autoround-a16w4 --local-dir Qwen3.5-4B-Autoround-a16w4
scp -r Qwen3.5-4B-Autoround-a16w4 sima@<modalix-ip>:/media/nvme/llima/models/

Replace <modalix-ip> with the IP address of your Modalix device.

Expected Directory Structure:

/media/nvme/llima/
└── models/
    └── Qwen3.5-4B-Autoround-a16w4/   # The compiled model

Usage

Validate with LLiMa CLI

Run the model directly on Modalix:

llima run Qwen3.5-4B-Autoround-a16w4

For all runtime options, run:

llima run -h

GenAI Demo Application

The GenAI demo application is separate from LLiMa installation. Use the GenAI Multimodal Assistant page to install and run the demo app. Once installed, the demo app can use precompiled models such as this one.

API Usage

To serve this model with OpenAI- or Ollama-compatible APIs and send requests to it, use the GenAI server workflow in Serve GenAI Models.

For direct VLM calls without setting up a server, see Run a VLM.

Recompilation

Use a Model Compiler and LLiMa build supporting Qwen3.5. The compiler versions recorded for these artifacts are in devkit/versions.txt.

The included compile.sh compiles the prequantized source checkpoint through the direct SiMa Builder path:

hf download simaai/Qwen3.5-4B-Autoround-Safetensors --local-dir Qwen3.5-4B-Autoround-Safetensors
bash compile.sh

The defaults match this build: 4096 context tokens, 128-token language groups, 448x448 images, filter sharing, and quantized embeddings and KV cache. Additional compiler arguments are forwarded to llima-compile.

8k context

To build for an 8,192-token context, recompile the matching source checkpoint with this repository's compile.sh in a compatible SiMa Model Compiler and LLiMa compilation environment. Change --max_num_tokens 4096 to --max_num_tokens 8192 in the script, then follow its source-checkpoint and output-directory instructions.

The context budget includes input and generated tokens. A larger context increases KV-cache memory requirements and must fit the target device. The downloadable artifacts use a 4,096-token context; any performance measurements in this card refer to that build.

Limitations

• Quantization: This model is quantized (A16W4) for optimal performance on embedded devices. Outputs may differ from the full-precision model.
• Fixed Resolution: While the standard Qwen3.5 architecture supports dynamic input resolutions, this version has been specifically optimized and fixed to 448x448 resolution at compile time to achieve maximum throughput and efficiency on the SiMa.ai MLA.

Troubleshooting

• sima-cli not found: Ensure that sima-cli is installed on your Modalix device.
• llima not found: Install or update the Neat Library. See Getting Started.
• Model can't be run: Verify the model directory is exactly inside /media/nvme/llima/models/ and not nested (e.g., /media/nvme/llima/models/Qwen3.5-4B-Autoround-a16w4/Qwen3.5-4B-Autoround-a16w4).
• Permission Denied: Ensure you have read/write permissions for the /media/nvme directory.

Resources

• GenAI with LLiMa
• Serve GenAI Models
• Run a VLM
• GenAI Multimodal Assistant