Gemma 3 27B IT on AMD Ryzen AI 9 HX PRO 370 / 375 NAS (48 GB)

Can the AMD Ryzen AI 9 HX PRO 370 / 375 NAS (48 GB) run Gemma 3 27B IT?

Yes, but slowly

Slow. At Q5_K_M it needs about 23.7 GB of 36.0 GB usable, with an estimated speed of 3.5 tok/s.

Best quant: Q5_K_M
Generation speed: 3.5 tok/s
Prompt processing: 651 tok/s
First token (1K prompt): 1.6 s
Largest safe context: 128K

Best app to run it on the AMD Ryzen AI 9 HX PRO 370 / 375 NAS (48 GB)

On Linux

Ollama

Command line + local API · runs on llama.cpp (ggml), MLX (Apple silicon, preview)

One command with Ollama’s own build of this model.
Serves an OpenAI-compatible API that other apps and coding tools can use.

ollama run gemma3:27b

llama.cpp Usually the first to support new model architectures.
KoboldCpp

Reverse engineering? Serve Gemma 3 27B IT 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, which apps run in a container on a NAS without a graphics card (models run on its processor), and whether the model fits fully in the memory models can use. Details.

Download Gemma 3 27B IT

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 gemma3:27b

Quantized files · GGUF

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

huggingface.co/ggml-org/gemma-3-27b-it-GGUF (opens in a new tab)

ollama run hf.co/ggml-org/gemma-3-27b-it-GGUF

Original weights · Hugging Face

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

huggingface.co/google/gemma-3-27b-it (opens in a new tab)

Free Hugging Face account needed: request access on the model page, and Google approves it (usually within minutes to a few hours).

Official repo · GitHub

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

github.com/google-deepmind/gemma (opens in a new tab)

Google DeepMind’s official Gemma library.

git clone https://github.com/google-deepmind/gemma

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

How it’s rated

Quality 69Good quality (B) · 69 of 100: how capable the model is at Q5_K_M, whatever it runs on. Match score 75/100 (quality, fit, speed and memory use) sets the ranking.: How capable the model itself is at Q5_K_M, 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 Slow: How well it runs on the AMD Ryzen AI 9 HX PRO 370 / 375 NAS (48 GB): 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.	31.4 / 45
Memory fitFull marks when the model fits in the share of the NAS’s memory that models can use (the rest stays with the NAS system and its storage services); fewer when it goes past that limit.	25.0 / 25
SpeedGeneration speed against a 40 tokens/s target for LLMs, or render time against a target for image and video.	4.4 / 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	75.8 / 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 Ryzen AI 9 HX PRO 370 / 375 NAS (48 GB)

Quant	Memory needed	Available	Speed	Runs	Quality	Download
`Q2_K`	15.2 GB	36.0 GB	6.9 tok/s	Slow	50Fair quality (C) · 50 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 63/100 (quality, fit, speed and memory use) sets the ranking.	GGUF files (opens in a new tab)
`Q3_K_M`	17.7 GB	36.0 GB	5.4 tok/s	Slow	60Good quality (B) · 60 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 69/100 (quality, fit, speed and memory use) sets the ranking.	GGUF files (opens in a new tab)
`Q4_K_M`	21.0 GB	36.0 GB	4.2 tok/s	Slow	68Good quality (B) · 68 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 74/100 (quality, fit, speed and memory use) sets the ranking.	GGUF (opens in a new tab)
`Q5_K_M`	23.7 GB	36.0 GB	3.5 tok/s	Slow	69Good quality (B) · 69 of 100: how capable the model is at Q5_K_M, whatever it runs on. Match score 75/100 (quality, fit, speed and memory use) sets the ranking.	GGUF files (opens in a new tab)
`Q6_K`	26.8 GB	36.0 GB	2.9 tok/s	Very slow	71Great quality (A) · 71 of 100: how capable the model is at Q6_K, whatever it runs on. Match score 76/100 (quality, fit, speed and memory use) sets the ranking.	GGUF files (opens in a new tab)
`Q8_0`	33.3 GB	36.0 GB	2.1 tok/s	Very slowtight fit	72Great quality (A) · 72 of 100: how capable the model is at Q8_0, whatever it runs on. Match score 74/100 (quality, fit, speed and memory use) sets the ranking.	GGUF (opens in a new tab)
`F16`	59.1 GB	36.0 GB	—	Won't run	72Great quality (A) · 72 of 100: how capable the model is at F16, whatever it runs on. Match score 9/100 (quality, fit, speed and memory use) sets the ranking.	GGUF · part 1 of 2 (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

71Great quality (A) · 71 of 100: how capable the model is at Q6_K, whatever it runs on. Match score 87/100 (quality, fit, speed and memory use) sets the ranking. NVIDIA GeForce RTX 5090 Q6_K · 55.7 tok/s · Runs great
68Good quality (B) · 68 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 85/100 (quality, fit, speed and memory use) sets the ranking. NVIDIA GeForce RTX 3090 Q4_K_M · 41.9 tok/s · Runs great
68Good quality (B) · 68 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 85/100 (quality, fit, speed and memory use) sets the ranking. NVIDIA GeForce RTX 4090 Q4_K_M · 45.2 tok/s · Runs great
68Good quality (B) · 68 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 85/100 (quality, fit, speed and memory use) sets the ranking. AMD Radeon RX 7900 XTX Q4_K_M · 37.0 tok/s · Runs great
60Good quality (B) · 60 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 81/100 (quality, fit, speed and memory use) sets the ranking. NVIDIA GeForce RTX 4080 SUPER Q3_K_M · 42.6 tok/s · Runs well (tight fit)
60Good quality (B) · 60 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 81/100 (quality, fit, speed and memory use) sets the ranking. NVIDIA GeForce RTX 5070 Ti Q3_K_M · 51.9 tok/s · Runs well (tight fit)

Quick answers

How much memory does Gemma 3 27B IT need on the AMD Ryzen AI 9 HX PRO 370 / 375 NAS 48GB?

At Q5_K_M it needs about 23.7 GB. The AMD Ryzen AI 9 HX PRO 370 / 375 NAS 48GB has 36.0 GB usable.

How fast is Gemma 3 27B IT on the AMD Ryzen AI 9 HX PRO 370 / 375 NAS 48GB?

About 3.5 tok/s at Q5_K_M, an estimate from the hardware’s memory bandwidth and the size of the model.

Which quantization of Gemma 3 27B IT should I use on the AMD Ryzen AI 9 HX PRO 370 / 375 NAS 48GB?

Q5_K_M. 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.