llama.cpp on the AMD Gorgon Halo 192GB
On the AMD Gorgon Halo 192GB, llama.cpp runs gpt-oss 117B · 5.1B active at 4-bit at a median 49.0 tokens per second for generation, from 2 plain runs (512 context, 1 not stated).
Speed by model on the AMD Gorgon Halo 192GB
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 |
|---|---|---|
| gpt-oss 117B · 5.1B active 4-bit | 49.0 t/smedian of 2, 512 context, 1 not stated | no other engine reported |
| DeepSeek V4 Flash 284B · 13B active 3-bit | 17.8 t/sone report, context not stated | no other engine reported |
| MiniMax M2.7 230B · 10B active 4-bit | 30.2 t/sone report, context not stated | no other engine reported |
| Qwen3-Coder 30B · 3B active 8-bit | 62.8 t/sone report, context not stated | no other engine reported |
| Qwen3.8 27B 8-bit | 8.1 t/sone report, context not stated | no other engine reported |
6 llama.cpp reports on the AMD Gorgon Halo 192GB
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.
- gpt-oss 117B · 5.1B active · MXFP455.7 t/s
- DeepSeek V4 Flash 284B · 13B active · UD-IQ3_S17.8 t/s
- MiniMax M2.7 230B · 10B active · UD-IQ4_XS30.2 t/s
- Qwen3.8 27B · Q8_08.1 t/s
- Qwen3-Coder 30B · 3B active · Q8_062.8 t/s
- gpt-oss 117B · 5.1B active · MXFP442.3 t/s
Frequently asked
How fast is llama.cpp on the AMD Gorgon Halo 192GB?
On the AMD Gorgon Halo 192GB, llama.cpp runs gpt-oss 117B · 5.1B active at 4-bit at a median 49.0 tokens per second for generation, from 2 plain runs (512 context, 1 not stated).