Can Hermes 4.3 36B run on RX 7900 XTX 24GB?

Q: What should I upgrade first if Hermes 4.3 36B feels slow on RX 7900 XTX 24GB?

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.

YES — With NVFP4

D39Poor

Estimated from fit model

Hermes 4.3 36B needs ~27.7 GB VRAM. RX 7900 XTX 24GB has 24.0 GB. With NVFP4 quantization, expect ~20 tok/s.

Runtime: llama.cppCapacity: OffloadBandwidth: HighStack: StandardBottleneck: Host offload

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.

Hermes 4.3 36B at Q4_K_M needs 29.5 GB — too much for RX 7900 XTX 24GB (24.0 GB). Runs at NVFP4 (27.7 GB) with medium quality. 3 quantization levels fit.

Capabilities:

Select quantization to explore

Q4_K_M (Medium quality) — 29.5 GB, exceeds 24.0 GB available

29.5 GB required24.0 GB available

123% VRAM needed

5.5 GB over capacity — needs offload or smaller quantization

Fit status

Too heavy

Decode

15.3 tok/s

TTFT

12643 ms

Safe context

Memory

29.5 GB / 24.0 GB

Offload

20%

Memory breakdown

Weights22.0 GB

KV Cache4.2 GB

Runtime0.9 GB

Headroom2.4 GB

See how fast it feels

With memory offload — actual speed may be lower

See how fast it feelsHermes 4.3 36B on RX 7900 XTX 24GB
1st promptCold start — includes initialization
>Solve: If a train travels at 60 mph for 2.5 hours, how far does it go?
Let me work through this step by step.

Given information:
  - Speed = 60 mph (miles per hour)
  - Time  = 2.5 hours

Formula: Distance = Speed x Time

Calculation:
  Distance = 60 mph x 2.5 h
           = 150 miles

The train travels 150 miles.

Double-check: at 60 mph the train covers 60 miles each hour. In half an hour it covers 30 miles. So in 2.5 hours: 60 + 60 + 30 = 150 miles. Confirmed.
2nd promptWarm — model ready, no init overhead
>What if the train stops for 20 minutes halfway?
The stop doesn't change the total distance covered while moving, but it affects total trip time.

Breakdown:
  - First half: 1.25 hours at 60 mph = 75 miles
  - Stop: 20 minutes (no distance)
  - Second half: 1.25 hours at 60 mph = 75 miles

Total distance: 150 miles (same)
Total time: 2.5 h + 0.33 h = 2.83 hours
Average speed: 150 / 2.83 = 53 mph

The distance stays the same but average speed drops to 53 mph because of the stop.
Estimated: 15.3 tok/s decode · 12.6s TTFT (warm) · 38 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 10% 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 2.7 GB of extra host RAM just for the offloaded portion, before OS and other tools.

Performance by workload

Workload	Grade	Fit	Decode	TTFT	Context
Chat	D	Very compromised (needs ~2.7 GB host RAM)	17.9 tok/s	5898 ms	4K
Coding	F	Too heavy	15.3 tok/s	12643 ms	4K
Agentic Coding	F	Too heavy	11.6 tok/s	24370 ms	4K
Reasoning	F	Too heavy	15.3 tok/s	14942 ms	4K
RAG	F	Too heavy	11.6 tok/s	30462 ms	4K

Inference speed

Hermes 4.3 36B inference speed — tokens per second by GPU & Mac

Estimated decode speed (tokens/sec) for Hermes 4.3 36B at Q4_K_M across popular GPUs and Apple Silicon, including multi-GPU rigs, using the fastest local runtime per device. Fastest is RTX 5090 32GB at ~55 tok/s. Speed is memory-bandwidth bound, so cards that fit the whole model in VRAM run far faster than ones that offload to system RAM.

GPU / Mac	Memory	Quant	Speed (tok/s)	Fits?
RTX 5090 32GB	32 GB	Q4_K_M	54.7	Offloads
2× RX 7900 XTX 24GB	48 GB	Q4_K_M	53.5	Fits

Quantization options

How Hermes 4.3 36B (36B params) fits at each quantization level on RX 7900 XTX 24GB (24.0 GB usable).

Quant	Bits	VRAM	Quality	Fit
Q2_K	2	14.0 GB	Low	C50
Q3_K_SBest for your GPU	3	17.6 GB	Low	C49

Get started

Copy-paste commands to run Hermes 4.3 36B on your machine.

Run

lms load hf-nousresearch--hermes-4-3-36b-gguf && lms server start

Upgrade options

Hardware that runs Hermes 4.3 36B well

Radeon AI PRO R9700 32GBBudget pick

32 GB VRAM (+8)

Makes the model fit on the accelerator instead of staying completely out of reach.17.2 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 32GBBest value

32 GB VRAM (+8)

Makes the model fit on the accelerator instead of staying completely out of reach.13.1 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

Radeon Pro W7800 32GBAMD upgrade

32 GB VRAM (+8)

Makes the model fit on the accelerator instead of staying completely out of reach.15.5 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.

~$2,499 MSRP

NVIDIA A100 40GBBiggest leap

40 GB VRAM (+16)1555 GB/s (+595)

Makes the model fit on the accelerator instead of staying completely out of reach.59.5 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

Can RX 7900 XTX 24GB run Hermes 4.3 36B?

Yes, RX 7900 XTX 24GB can run Hermes 4.3 36B at NVFP4 quantization (Very compromised (needs ~2.7 GB host RAM)). The recommended Q4_K_M requires 29.5 GB which exceeds available memory, but at NVFP4 it needs only 27.7 GB. Expected decode speed: 20.0 tok/s.

How much VRAM does Hermes 4.3 36B need?

Hermes 4.3 36B (36B parameters) requires approximately 29.5 GB at Q4_K_M quantization. On RX 7900 XTX 24GB, it fits at NVFP4 using 27.7 GB.

What is the best quantization for Hermes 4.3 36B?

The recommended quantization is Q4_K_M, but on RX 7900 XTX 24GB the best fitting quantization is NVFP4, which uses 27.7 GB.

What speed will Hermes 4.3 36B run at on RX 7900 XTX 24GB?

On RX 7900 XTX 24GB, Hermes 4.3 36B achieves approximately 20.0 tokens per second decode speed with a time-to-first-token of 9681ms using NVFP4 quantization.

Can RX 7900 XTX 24GB run Hermes 4.3 36B for coding?

For coding workloads, Hermes 4.3 36B on RX 7900 XTX 24GB receives a F grade with 15.3 tok/s and 4K context.

What context window can Hermes 4.3 36B use on RX 7900 XTX 24GB?

On RX 7900 XTX 24GB, Hermes 4.3 36B can safely use up to 4K tokens of context at NVFP4 quantization. The model's official context limit is —, but available memory constrains the safe maximum.

See all results for RX 7900 XTX 24GB See all hardware for Hermes 4.3 36B

Embed this result▼

Paste this snippet into any page to show a live fit card.

<iframe src="https://willitrunai.com/embed/hf-nousresearch--hermes-4-3-36b-gguf-on-rx-7900-xtx-24gb" width="400" height="180" frameborder="0" style="border:none;border-radius:12px;overflow:hidden;" title="Will It Run AI — fit result"></iframe>

Preview: