llama.cpp on the AMD RX 7900 XTX
On the AMD RX 7900 XTX, llama.cpp runs Gemma 4 26B · 4B active at 4-bit at a median 142.8 tokens per second for generation, from 1 plain run (context not stated).
Speed by model on the AMD RX 7900 XTX
Plain runs only: one device, one request, no speculative decoding, the whole model in memory. Each row is one model size at one bit class, with the other engines' plain runs of the same model on this card beside it. Context length and build still differ between runs, and each side shows its own.
| Model | llama.cpp | Other engines |
|---|---|---|
| Gemma 4 26B · 4B active 4-bit | 142.8 t/sone report, context not stated | no other engine reported |
| Gemma 4 31B 4-bit | 30.1 t/sone report, 32K context | no other engine reported |
16 llama.cpp reports on the AMD RX 7900 XTX
Newest first. A run on several cards, with speculative decoding, with part of the model in system RAM or with batched requests says so, since each describes a different setup.
- Qwen3.8 125B · 6B active · UD_Q3_XXLPP 920.0 t/s(2 cards, speculative decoding, part in system RAM)
- Qwen3.8 27B · Q877.0 t/s(4 cards, speculative decoding)
- Qwen3.8 27B · IQ3_S44.5 t/s(speculative decoding)
- Qwen3.8 27B · IQ3_S68.9 t/s(speculative decoding)
- Qwen3.8 27B23.8 t/s
- Qwen3.8 · IU423.7 t/s(2 cards, speculative decoding)
- Model not identified · Q4_0126.1 t/s
- Gemma 4 26B · 4B active · Q4_0142.8 t/s
- Gemma 4 31B · Q4_K_XL30.1 t/s
- Qwen3.8 · Q3_K_XL24.0 t/s(2 cards, speculative decoding, part in system RAM)
- DeepSeek V4 Flash 284B · 13B active · UD-IQ3_XXS12.0 t/s(4 cards)
- Qwen3.8 27B · Q4_K_M72.1 t/s
- Qwen3.8 27B · Q4_K_M25.0 t/s(speculative decoding)
- Qwen3.8 27B · Q6_K_XL27.0 t/s(2 cards, speculative decoding)
- Qwen3.8 125B · 6B active · UD-Q4_K_XL14.4 t/s(part in system RAM)
- Qwen3.811.0 t/s
Frequently asked
How fast is llama.cpp on the AMD RX 7900 XTX?
On the AMD RX 7900 XTX, llama.cpp runs Gemma 4 26B · 4B active at 4-bit at a median 142.8 tokens per second for generation, from 1 plain run (context not stated).