VOOZH about

URL: https://willitrunai.com/can-run/exaone-4-32b-on-rx-9070-16gb

⇱

Can EXAONE 4.0 32B run on RX 9070 16GB?

YES — With Q2_K

B67Good
○Estimated from fit model

EXAONE 4.0 32B needs ~18.9 GB VRAM. RX 9070 16GB has 16.0 GB. With Q2_K quantization, expect ~16 tok/s.

Runtime: llama.cppCapacity: OffloadBandwidth: MediumStack: StandardBottleneck: Host offload
Share:

Operating mode

Choose the run profile you care about

Interactive favors responsiveness, while light API and scale-out lean harder on serving readiness. The fit stays the same, but the recommendation lens changes.

Current mode

Balanced

Balanced for general local use. Keeps the ranking neutral across personal and serving workflows.

EXAONE 4.0 32B at Q4_K_M needs 25.9 GB — too much for RX 9070 16GB (16.0 GB). Runs at Q2_K (18.9 GB) with low quality.
Capabilities:

Select quantization to explore

Q4_K_M (Medium quality) — 25.9 GB, exceeds 16.0 GB available
25.9 GB required16.0 GB available
162% VRAM needed

9.9 GB over capacity — needs offload or smaller quantization

Fit status

Too heavy

Decode

6.3 tok/s

TTFT

30869 ms

Safe context

4K

Memory

25.9 GB / 16.0 GB

Offload

40%

Memory breakdown

Weights19.5 GB
KV Cache3.9 GB
Runtime0.9 GB
Headroom1.6 GB

See how fast it feels

With memory offload — actual speed may be lower
See how fast it feelsEXAONE 4.0 32B on RX 9070 16GB
1st promptCold start — includes initialization
>Write a Python function to sort a list
Here's a Python function that sorts a list using quicksort: def quicksort(arr): if len(arr) <= 1: return arr pivot = arr[len(arr) // 2] left = [x for x in arr if x < pivot] middle = [x for x in arr if x == pivot] right = [x for x in arr if x > pivot] return quicksort(left) + middle + quicksort(right) # Example usage numbers = [3, 6, 8, 10, 1, 2, 1] print(quicksort(numbers)) # [1, 1, 2, 3, 6, 8, 10] The algorithm picks a pivot, partitions the list into smaller, equal, and larger elements, then recursively sorts each partition and concatenates the results.
2nd promptWarm — model ready, no init overhead
>Now add type hints and a docstring
def quicksort(arr: list[int]) -> list[int]: """Sort a list using the quicksort algorithm. Args: arr: List of integers to sort. Returns: A new sorted list. """ if len(arr) <= 1: return arr pivot = arr[len(arr) // 2] left = [x for x in arr if x < pivot] middle = [x for x in arr if x == pivot] right = [x for x in arr if x > pivot] return quicksort(left) + middle + quicksort(right) Added type hints for the input and return type, plus a Google-style docstring.
Estimated: 6.3 tok/s decode · 30.9s TTFT (warm) · 16 tok/s prefill

What limits this setup

It fits through host-memory offload, and offload is the main reason performance drops.

CPU or host-memory offload is active

About 20% of the working set spills out of accelerator memory, which usually hurts latency and sustained decode throughput.

Very little memory headroom

You can run the model, but there is not much room left for longer context, bigger batches, extra apps, or future model updates.

Best improvement path

Remove offload with more accelerator memory

Prioritize a GPU or unified-memory tier that fits the whole model natively. Removing offload usually helps more than small compute gains.

Buy headroom, not only minimum fit

A slightly larger memory tier gives you safer context growth and makes the recommendation more future-proof.

Increase host RAM if you keep offloading

This setup may need roughly 1.9 GB of extra host RAM just for the offloaded portion, before OS and other tools.

Performance by workload

WorkloadGradeFitDecodeTTFTContext
ChatFToo heavy6.8 tok/s15499 ms4K
CodingFToo heavy5.8 tok/s33338 ms4K
Agentic CodingFToo heavy4.4 tok/s64567 ms4K
ReasoningFToo heavy5.8 tok/s39400 ms4K
RAGFToo heavy4.4 tok/s80708 ms4K

Quantization options

How EXAONE 4.0 32B (32B params) fits at each quantization level on RX 9070 16GB (16.0 GB usable).

QuantBitsVRAMQualityFit
Q2_K
2
12.5 GB
LowF0
Q3_K_S
3
15.7 GB
LowF0
NVFP4
4

Get started

Copy-paste commands to run EXAONE 4.0 32B on your machine.

Run

ollama run exaone-4:32b

Upgrade options

Hardware that runs EXAONE 4.0 32B well

RX 7900 XTX 24GBBest value
24 GB VRAM (+8)960 GB/s (+320)
A
Makes the model fit on the accelerator instead of staying completely out of reach.22.9 tok/s decode

Makes the model fit on the accelerator instead of staying completely out of reach.

Raises estimated decode speed by about 263%.

~$999 MSRP

Radeon AI PRO R9700 32GBBudget pick
32 GB VRAM (+16)
A
Makes the model fit on the accelerator instead of staying completely out of reach.20.9 tok/s decode

Makes the model fit on the accelerator instead of staying completely out of reach.

Removes host-memory offload, which is usually the single biggest latency and throughput win.

~$1,899 MSRP

Radeon Pro W6800 32GBAMD upgrade
32 GB VRAM (+16)
A
Makes the model fit on the accelerator instead of staying completely out of reach.15.9 tok/s decode

Makes the model fit on the accelerator instead of staying completely out of reach.

Removes host-memory offload, which is usually the single biggest latency and throughput win.

~$2,249 MSRP

👁 NVIDIA
NVIDIA A100 40GBBiggest leap
40 GB VRAM (+24)1555 GB/s (+915)
S
Makes the model fit on the accelerator instead of staying completely out of reach.72.3 tok/s decode

Makes the model fit on the accelerator instead of staying completely out of reach.

Removes host-memory offload, which is usually the single biggest latency and throughput win.

~$10,000 MSRP

Frequently asked questions

See all results for RX 9070 16GBSee all hardware for EXAONE 4.0 32B
17.9 GB
Medium
F0
Q4_K_M
4
19.5 GB
MediumF0
Q5_K_M
5
23.0 GB
HighF0
Q6_K
6
26.2 GB
HighF0
Q8_0
8
34.2 GB
Very HighF0
F16
16
65.6 GB
MaximumF0

Remove offload with more accelerator memory. Prioritize a GPU or unified-memory tier that fits the whole model natively. Removing offload usually helps more than small compute gains.