Qwen3.5 35B A3B (Q3_K_M) — 26.8 GBon NVIDIA RTX A6000 384GB

Q3_K_M NVIDIA RTX A6000 384GB

Overview

Qwen3.5 35B A3B is a 35.95B parameter moe language model by Qwen, with code, multilingual, thinking, tool-calls, vision capabilities. It supports a context window of up to 262,144 tokens.

Qwen3.5 35B A3B is a Mixture-of-Experts model from Alibaba's Qwen team with 35 billion total parameters but only 3 billion active per token, routed across 256 experts for extreme efficiency. It is natively multimodal, processing text, images, and video, with built-in thinking capabilities for chain-of-thought reasoning. The model supports a 262K context window and covers over 200 languages. Released under the Apache 2.0 license, it delivers flagship-level performance at a fraction of the compute cost, quantizing efficiently for self-hosted deployment on consumer hardware.

At Q3_K_M quantization (low quality tier), the model weighs 15.54 GB. This fits within the 384 GB of VRAM on NVIDIA RTX A6000 384GB, enabling full GPU inference.

Hardware Requirements

Model size	15.54 GB
VRAM available	384 GB
VRAM used	26.8 GB
GPU layers	40 / 40
Context size	262,144
Backend	cuda13
Flash attention	Yes

Performance Notes

Mixture-of-Experts

This is a Mixture-of-Experts model with 256 total experts, of which 8 are active per token. Only the active experts need fast GPU access — inactive expert weights can reside in system RAM with minimal impact on generation speed.

Read the MoE deployment guide →

Deploy

Server Local

Prerequisites

Ensure your GPU nodes are prepared with the NVIDIA container toolkit:

ansible-playbook prositronic.infra.nvidia_container_toolkit

Command

helmfile --state-values-file <(curl -s https://www.prositronic.eu/values/qwen3-5-35b-a3b/q3_k_m/nvidia-a6000-384gb.yaml) apply

Generated values.yaml

/values/qwen3-5-35b-a3b/q3_k_m/nvidia-a6000-384gb.yaml

Loading values…

Install llama.cpp

Install llama.cpp from the official build instructions for your platform and backend.

Download Model

curl -L -o qwen3-5-35b-a3b.gguf "https://huggingface.co/unsloth/Qwen3.5-35B-A3B-GGUF/resolve/main/Qwen3.5-35B-A3B-UD-Q3_K_M.gguf"

Start Server

llama-server \
  -m qwen3-5-35b-a3b.gguf \
  --n-gpu-layers 40 \
  --ctx-size 262144 \
  --flash-attn

Verify

curl http://localhost:8080/health

Explore

• Configure a deployment
• Browse models
• Browse hardware

Other quantizations

• Qwen3.5 35B A3B (Q4_K_M)

Other hardware

• Qwen3.5 35B A3B on CPU Only
• Qwen3.5 35B A3B on NVIDIA A100 80GB
• Qwen3.5 35B A3B on NVIDIA H100

Frequently Asked Questions

How much VRAM does Qwen3.5 35B A3B (Q3_K_M) need?

The Q3_K_M quantization of Qwen3.5 35B A3B requires 15.54 GB. All 40 layers fit in the 384 GB of VRAM available on NVIDIA RTX A6000 384GB, enabling full GPU acceleration.

Can I run Qwen3.5 35B A3B on NVIDIA RTX A6000 384GB?

Yes. NVIDIA RTX A6000 384GB provides 384 GB of VRAM, which is enough to run Qwen3.5 35B A3B (Q3_K_M) with all layers on the GPU for optimal performance.

What is quantization?

Quantization reduces a model's numerical precision from its original floating-point format to a more compact representation. This shrinks the file size and VRAM footprint, making it possible to run large models on consumer hardware. The trade-off is a small reduction in output quality. Q3_K_M compresses Qwen3.5 35B A3B from its original size down to 15.54 GB.

What quantization should I choose for Qwen3.5 35B A3B?

Q3_K_M is a low-quality quantization. Higher-quality quants (Q8, Q6) preserve more model accuracy but need more VRAM. Lower quants (Q4, Q3, Q2) reduce VRAM usage at the cost of some quality. Choose based on your available hardware and quality requirements.

What is flash attention and why is it enabled?

Flash attention is a memory-efficient algorithm that speeds up the attention mechanism in transformer models. It reduces VRAM usage during inference by avoiding the materialisation of the full attention matrix. For Qwen3.5 35B A3B on NVIDIA RTX A6000 384GB, flash attention is enabled to maximise context length and throughput within the available 384 GB of VRAM.

What is MoE and how does it affect deployment?

Qwen3.5 35B A3B uses a Mixture-of-Experts (MoE) architecture with 256 experts, of which 8 are active per token. This means only a fraction of the model weights are used for each inference step, allowing MoE models to be larger in total parameter count while remaining efficient at inference time.

How do I run Qwen3.5 35B A3B (Q3_K_M) with Ollama?

Run ollama run qwen3.5:35b-a3b-q3_k_m to start Qwen3.5 35B A3B (Q3_K_M). Ollama handles downloading the model weights automatically on first run.

Last updated: March 14, 2026