Step 3.5 Flash (196B-A11B) on AMD Instinct MI350X

Can the AMD Instinct MI350X run Step 3.5 Flash (196B-A11B)?

Yes, it runs great

Runs great. At Q8_0 it needs about 227 GB of 288 GB usable, with an estimated speed of 157 tok/s.

Best quant: Q8_0
Generation speed: 157 tok/s
Prompt processing: 7,722 tok/s
First token (1K prompt): 0.3 s
Largest safe context: 256K

Best app to run it on the AMD Instinct MI350X

On Linux

vLLM

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 stepfun-ai/Step-3.5-Flash

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 bartowski/stepfun-ai_Step-3.5-Flash-GGUF:Q8_0

LM Studio Point-and-click: search, download and chat in one window.
llama.cpp Usually the first to support new model architectures.

Want an agent? Step 3.5 Flash (196B-A11B) supports tool use, so you can pair a llama.cpp or MLX LM server with one of these. Apps that run on this computer come first.

Coding agents

Personal assistants

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 Step 3.5 Flash (196B-A11B)

Recommended for the AMD Instinct MI350X: Q8_0. Every other quantization has its own download link in the table below.

Quantized files · GGUF

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

Download the Q8_0 files (opens in a new tab)

huggingface.co/bartowski/stepfun-ai_Step-3.5-Flash-GGUF (opens in a new tab)

ollama run hf.co/bartowski/stepfun-ai_Step-3.5-Flash-GGUF

Original weights · Hugging Face

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

huggingface.co/stepfun-ai/Step-3.5-Flash (opens in a new tab)

No sign-in needed.

Official repo · GitHub

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

github.com/stepfun-ai/Step-3.5-Flash (opens in a new tab)

git clone https://github.com/stepfun-ai/Step-3.5-Flash

Links checked 2 days ago against Hugging Face, Ollama and GitHub.

How it’s rated

Quality 92Excellent quality (A+) · 92 of 100: how capable the model is at Q8_0, whatever it runs on. Match score 96/100 (quality, fit, speed and memory use) sets the ranking.: How capable the model itself is at Q8_0, 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 MI350X: 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.	41.4 / 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.	15.0 / 15
Match score	96.4 / 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 MI350X

Quant	Memory needed	Available	Speed	Runs	Quality	Download
`Q2_K`	95.2 GB	288 GB	255 tok/s	Runs great	64Good quality (B) · 64 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 76/100 (quality, fit, speed and memory use) sets the ranking.	GGUF · folder (opens in a new tab)
`Q3_K_M`	113 GB	288 GB	235 tok/s	Runs great	77Great quality (A) · 77 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 83/100 (quality, fit, speed and memory use) sets the ranking.	GGUF · folder (opens in a new tab)
`Q4_K_M`	137 GB	288 GB	214 tok/s	Runs great	87Excellent quality (A+) · 87 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 90/100 (quality, fit, speed and memory use) sets the ranking.	GGUF · folder (opens in a new tab)
`Q5_K_M`	157 GB	288 GB	197 tok/s	Runs great	89Excellent quality (A+) · 89 of 100: how capable the model is at Q5_K_M, whatever it runs on. Match score 92/100 (quality, fit, speed and memory use) sets the ranking.	GGUF · folder (opens in a new tab)
`Q6_K`	179 GB	288 GB	182 tok/s	Runs great	91Excellent quality (A+) · 91 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.	GGUF · folder (opens in a new tab)
`Q8_0`	227 GB	288 GB	157 tok/s	Runs great	92Excellent quality (A+) · 92 of 100: how capable the model is at Q8_0, whatever it runs on. Match score 96/100 (quality, fit, speed and memory use) sets the ranking.	GGUF · folder (opens in a new tab)
`F16`	414 GB	288 GB	—	Won't run	92Excellent quality (A+) · 92 of 100: how capable the model is at F16, whatever it runs on. Match score 12/100 (quality, fit, speed and memory use) sets the ranking.	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

89Excellent quality (A+) · 89 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 M3 Ultra (256 GB) Q5_K_M · 53.0 tok/s · Runs great
87Excellent quality (A+) · 87 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 93/100 (quality, fit, speed and memory use) sets the ranking. Apple M2 Ultra (192 GB) Q4_K_M · 61.1 tok/s · Runs well (tight fit)
91Excellent quality (A+) · 91 of 100: how capable the model is at Q6_K, whatever it runs on. Match score 92/100 (quality, fit, speed and memory use) sets the ranking. Apple M3 Ultra (512 GB) Q6_K · 45.6 tok/s · Runs great
77Great quality (A) · 77 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 87/100 (quality, fit, speed and memory use) sets the ranking. NVIDIA GB10 (128 GB) Q3_K_M · 34.9 tok/s · Runs well (tight fit)
64Good quality (B) · 64 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 82/100 (quality, fit, speed and memory use) sets the ranking. Apple M4 Max (128 GB) Q2_K · 67.9 tok/s · Runs well (tight fit)
64Good quality (B) · 64 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 82/100 (quality, fit, speed and memory use) sets the ranking. NVIDIA RTX PRO 6000 Blackwell Q2_K · 162 tok/s · Runs well (tight fit)

Quick answers

How much memory does Step 3.5 Flash (196B-A11B) need on the AMD Instinct MI350X 288GB?

At Q8_0 it needs about 227 GB. The AMD Instinct MI350X 288GB has 288 GB usable.

How fast is Step 3.5 Flash (196B-A11B) on the AMD Instinct MI350X 288GB?

About 157 tok/s at Q8_0, an estimate from the hardware’s memory bandwidth and the size of the model.

Which quantization of Step 3.5 Flash (196B-A11B) should I use on the AMD Instinct MI350X 288GB?

Q8_0. 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.