Can the AMD Instinct MI300X run Qwen3.5 35B-A3B?

Yes, it runs great

Runs great. At F16 it needs about 82.3 GB of 192 GB usable, with an estimated speed of 150 tok/s.

Best quant: F16
Generation speed: 150 tok/s
Prompt processing: 33,769 tok/s
First token (1K prompt): 0.3 s
Largest safe context: 256K

Best app to run it on the AMD Instinct MI300X

On Linux

Inference server · runs on vLLM

The full model fits in GPU memory at 8-bit, so it can batch many requests: best for serving a team or an app.

vllm serve Qwen/Qwen3.5-35B-A3B

Lemonade AMD’s own app: llama.cpp tuned for Radeon on Vulkan and ROCm.
LM Studio Point-and-click: search, download and chat in one window.

On Windows

Lemonade

Desktop app · runs on llama.cpp, ONNX Runtime GenAI (Ryzen AI NPU)

AMD’s own app: llama.cpp tuned for Radeon on Vulkan and ROCm.

lemonade pull unsloth/Qwen3.5-35B-A3B-GGUF:BF16

LM Studio Point-and-click: search, download and chat in one window.
Ollama One command with Ollama’s own build of this model.

Reverse engineering? Serve Qwen3.5 35B-A3B locally and use it from Ghidra, IDA Pro, Binary Ninja, radare2 or JADX with one of these:

How we choose: the files published for this model, your operating system and GPU, whether it fits fully in memory, and published speed tests (MLX vs llama.cpp on Apple silicon, vLLM vs llama.cpp vs Ollama on NVIDIA). Details.

Download Qwen3.5 35B-A3B

Recommended for the AMD Instinct MI300X: F16. Every other quantization has its own download link in the table below.

Quickest · Ollama

One command downloads a ready-to-run version and starts a chat.

Get Ollama at ollama.com/download (opens in a new tab) · model page (opens in a new tab)

ollama run qwen3.5:35b-a3b-bf16

Quantized files · GGUF

Works in LM Studio, llama.cpp, Jan and Ollama. No sign-in needed.

Download the F16 files (opens in a new tab)

huggingface.co/unsloth/Qwen3.5-35B-A3B-GGUF (opens in a new tab)

ollama run hf.co/unsloth/Qwen3.5-35B-A3B-GGUF

Original weights · Hugging Face

Full-precision files from Alibaba, for fine-tuning, vLLM, Transformers.

huggingface.co/Qwen/Qwen3.5-35B-A3B (opens in a new tab)

No sign-in needed.

Official repo · GitHub

Reference code, inference scripts and docs from Alibaba. The weights themselves download from Hugging Face.

github.com/QwenLM/Qwen3.8 (opens in a new tab)

Qwen’s current repo: Qwen3.5 and Qwen3.6 redirect here.

git clone https://github.com/QwenLM/Qwen3.8

Links checked 3 days ago against Hugging Face, Ollama and GitHub.

How it’s rated

Quality 93Excellent quality (A+) · 93 of 100: how capable the model is at F16, whatever it runs on. Match score 91/100 (quality, fit, speed and memory use) sets the ranking.: How capable the model itself is at F16, from its editorial quality tier scaled by the quantization or precision. It’s the same on any hardware that runs this variant. Scale: A+ 85+ · A 70+ · B 55+ · C 40+ · D under 40.
Runs Runs great: How well it runs on the AMD Instinct MI300X: speed, capped by how well it fits in memory. Runs great: 30+ tokens/s (or near the image, video or decision speed target) · Runs well: 10+ tokens/s · Slow: 3+ tokens/s · Very slow: under 3 · Won’t run: doesn’t fit even with offload. A tight fit or RAM offload caps it at Runs well; heavy offload caps it at Slow.

Match score

To order lists, each result also gets a match score out of 100 from four parts. It favors the most capable model that still runs well.

Model qualityEditorial quality tier of the model, scaled by the quantization or precision used.	41.9 / 45
Memory fitFull marks when everything sits in GPU memory; fewer when part spills to system RAM.	25.0 / 25
SpeedGeneration speed against a 40 tokens/s target for LLMs, or render time against a target for image and video.	15.0 / 15
Memory useRewards using the memory you have, so a bigger, better variant beats a tiny one that leaves most memory idle.	9.9 / 15
Match score	91.7 / 100

The match score orders results; it is not shown as a grade. Quality and Runs are the two ratings in the tables.

Every quantization on the AMD Instinct MI300X

Quant	Memory needed	Available	Speed	Runs	Quality	Download
`Q2_K`	24.7 GB	192 GB	290 tok/s	Runs great	65Good quality (B) · 65 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 72/100 (quality, fit, speed and memory use) sets the ranking.	GGUF (UD-Q2_K_XL) (opens in a new tab)
`Q3_K_M`	28.0 GB	192 GB	274 tok/s	Runs great	78Great quality (A) · 78 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 78/100 (quality, fit, speed and memory use) sets the ranking.	GGUF (opens in a new tab)
`Q4_K_M`	32.3 GB	192 GB	256 tok/s	Runs great	88Excellent quality (A+) · 88 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 83/100 (quality, fit, speed and memory use) sets the ranking.	GGUF (opens in a new tab)
`Q5_K_M`	35.9 GB	192 GB	241 tok/s	Runs great	90Excellent quality (A+) · 90 of 100: how capable the model is at Q5_K_M, whatever it runs on. Match score 84/100 (quality, fit, speed and memory use) sets the ranking.	GGUF (opens in a new tab)
`Q6_K`	39.8 GB	192 GB	228 tok/s	Runs great	92Excellent quality (A+) · 92 of 100: how capable the model is at Q6_K, whatever it runs on. Match score 86/100 (quality, fit, speed and memory use) sets the ranking.	GGUF (opens in a new tab)
`Q8_0`	48.5 GB	192 GB	206 tok/s	Runs great	93Excellent quality (A+) · 93 of 100: how capable the model is at Q8_0, whatever it runs on. Match score 87/100 (quality, fit, speed and memory use) sets the ranking.	GGUF (opens in a new tab)
`F16`	82.3 GB	192 GB	150 tok/s	Runs great	93Excellent quality (A+) · 93 of 100: how capable the model is at F16, whatever it runs on. Match score 91/100 (quality, fit, speed and memory use) sets the ranking.	GGUF · folder (opens in a new tab) Original weights (opens in a new tab)

Memory figures assume 8K of context. Longer conversations or documents grow the KV cache.

Other hardware that runs it

93Excellent quality (A+) · 93 of 100: how capable the model is at F16, whatever it runs on. Match score 96/100 (quality, fit, speed and memory use) sets the ranking. NVIDIA RTX PRO 6000 Blackwell F16 · 98.2 tok/s · Runs great
92Excellent quality (A+) · 92 of 100: how capable the model is at Q6_K, whatever it runs on. Match score 96/100 (quality, fit, speed and memory use) sets the ranking. Apple M1 Max (64 GB) Q6_K · 72.3 tok/s · Runs great
92Excellent quality (A+) · 92 of 100: how capable the model is at Q6_K, whatever it runs on. Match score 96/100 (quality, fit, speed and memory use) sets the ranking. Apple M4 Max (64 GB) Q6_K · 90.1 tok/s · Runs great
90Excellent quality (A+) · 90 of 100: how capable the model is at Q5_K_M, whatever it runs on. Match score 95/100 (quality, fit, speed and memory use) sets the ranking. Apple M4 Pro (48 GB) Q5_K_M · 62.2 tok/s · Runs great
92Excellent quality (A+) · 92 of 100: how capable the model is at Q6_K, whatever it runs on. Match score 95/100 (quality, fit, speed and memory use) sets the ranking. NVIDIA GeForce RTX 5090 Q6_K · 179 tok/s · Runs well (tight fit)
93Excellent quality (A+) · 93 of 100: how capable the model is at F16, whatever it runs on. Match score 95/100 (quality, fit, speed and memory use) sets the ranking. Apple M2 Ultra (192 GB) F16 · 51.3 tok/s · Runs great

Quick answers

How much memory does Qwen3.5 35B-A3B need on the AMD Instinct MI300X 192GB?

At F16 it needs about 82.3 GB. The AMD Instinct MI300X 192GB has 192 GB usable.

How fast is Qwen3.5 35B-A3B on the AMD Instinct MI300X 192GB?

About 150 tok/s at F16, an estimate from the hardware’s memory bandwidth and the size of the model.

Which quantization of Qwen3.5 35B-A3B should I use on the AMD Instinct MI300X 192GB?

F16. Of the versions that fit, it is the one that runs best; when two run equally well, the check prefers the higher-rated one.

Estimates come from published hardware specs and our open formulas, not from running the model on your machine. Real speed varies with runtime, drivers and settings. Numbers marked “Measured” come from a linked benchmark.