Can Hermes 4.3 36B run on the RX 7800 XT?
Not fully in VRAM. Even at 2-bit with 8K context, Hermes 4.3 36B needs about 15.4 GB, more than the 15.2 GB usable on the RX 7800 XT. It can still run with offload (slower): Q4_K_M with part of the model in system RAM, about 2.6–5.0 tokens/sec (estimate). Details and command.
Runs with offload (slower)
Runs with offload 4-bit (Q4_K_M) at 8K context: 40 of 64 layers on the RX 7800 XT (about 14.6 GB) and the other 24 in system RAM (about 8.4 GB). Needs a PC with at least 32 GB of system RAM. Estimated speed: 2.6–5.0 tokens/sec (slow but usable), versus a bandwidth ceiling of ~29 tok/s if it fit fully in VRAM.
llama-server -hf bartowski/NousResearch_Hermes-4.3-36B-GGUF:Q4_K_M -c 8192 -ngl 40
Estimate, not a benchmark. The command downloads the GGUF from Hugging Face. The speed range spans dual-channel DDR4-3200 to DDR5-6000 system RAM; your CPU, RAM speed and llama.cpp version move it. If llama.cpp runs out of GPU memory, lower -ngl by 2-3; if VRAM is left over, raise it for more speed. Recent llama.cpp builds can also pick the split automatically. How we estimate offload.
Memory by quant and context
| Quant | 4K ctx | 8K ctx | 32K ctx | 128K ctx | Max tok/s (est., 8K) |
|---|---|---|---|---|---|
| 2-bit (Q2_K) • | 14.4 GB Tight | 15.4 GB No | 21.4 GB No | 45.4 GB No | — |
| 3-bit (Q3_K_M) • | 18.1 GB No | 19.1 GB No | 25.1 GB No | 49.1 GB No | — |
| 4-bit (Q4_K_M) • | 22.0 GB No | 23.0 GB No | 29.0 GB No | 53.0 GB No | — |
| 5-bit (Q5_K_M) • | 25.5 GB No | 26.5 GB No | 32.5 GB No | 56.5 GB No | — |
| 6-bit (Q6_K) • | 29.3 GB No | 30.3 GB No | 36.3 GB No | 60.3 GB No | — |
| 8-bit (Q8_0) • | 37.5 GB No | 38.5 GB No | 44.5 GB No | 68.5 GB No | — |
| 16-bit (F16/BF16) • | 69.0 GB No | 70.0 GB No | 76.0 GB No | 100 GB No | — |
• real GGUF file size from bartowski/NousResearch_Hermes-4.3-36B-GGUF. Speed is a bandwidth ceiling (624 GB/s), not a benchmark. Methodology.
Hardware that does fit it
RX 7900 XTRTX 3090RTX 3090 TiRTX 4090RX 7900 XTXMac M2 24GB
What fits instead
- Cheapest listed GPU that runs Hermes 4.3 36B: RTX 5090 (32 GB, launch MSRP $1,999) at Q5_K_M, fully in VRAM at 8K context.
- Largest popular model that fits the RX 7800 XT: gpt-oss 20B (21B, MoE) at Q8_0, about 12.3 GB at 8K context.
"Runs" and "fits" mean the whole model in GPU memory at 8K context (4-bit or better for the model pick). Prices are US launch MSRPs, not today's street prices. "Popular" means the 60 most-downloaded models we track.
More
Hermes 4.3 36B on every GPU and Mac · Every model that runs on the RX 7800 XT · Other models that fit 16 GB VRAM · Try other settings in the calculator
Guides: Best GPU for local LLMs by budget · Can I run gpt-oss, Llama 70B or Qwen3? · Mac vs PC for local LLMs
Data: Hugging Face Hub (NousResearch/Hermes-4.3-36B), AMD specs. Model data last checked 2026-11-17; refreshed hourly.