Machine
INTEL Arc B390 · Core Ultra X7 358H #a650b289
86 runs of 13 models since Oct 8, 2026. Last run Oct 12, 2026.
86
runs
13
models
59
configs
27
configs run 2+ times
Hardware
- GPU
- Intel Arc B390
- CPU
- Intel(R) Core(TM) Ultra X7 358H 16 cores
- RAM
- 30 GB
- OS
- CachyOS
- Tools
- llama.cpp 72 · Ollama 14
- Client
- 0.4.89
Models on this machine
One dot per model, at its best config. Up is better answers, right is faster. Time to finish counts every token the model wrote, thinking included, so a model that talks a lot ends up further left.
Best trade-off: quality 60+, and nothing else here is both faster and betterOther models · dot size = number of runs
Medians of each model's best config, 13 models.
Best config per model
sort
Runs with the same model, quant, tool, context, thinking and KV cache count as one config. Numbers are the median of its runs, and each dot is one run on the 0–100 quality scale. ★ marks the best config with at least 2 rated runs.
configrunsqualityeach runspeed
qwen3-coder
3 runs · 3 configs
unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF
9 runs · 5 configs
★ bestllama.cpp32k ctxthinking default
×2 runs
65.661.9–69.4
steady
18.7 tok/s
13.8–23.6
qwen3.5
4 runs · 4 configs
yuxinlu1/gemma-4-12B-agentic-fable5-composer2.5-v2-3.5x-tau2-GGUF
9 runs · 5 configs
★ bestllama.cpp16k ctxthinking default
×2 runs
62.059.0–65.0
steady
7.1 tok/s
6.7–7.4
richardyoung/Qwen3-14B-abliterated-GGUF
8 runs · 5 configs
★ bestllama.cpp16k ctxthinking default
×2 runs
58.051.4–64.7
varies
6.9 tok/s
6.1–7.7
gemma3
4 runs · 4 configs
Qwen/Qwen2.5-Coder-14B-Instruct-GGUF
9 runs · 5 configs
★ bestllama.cpp66k ctxthinking default
×2 runs
54.046.7–61.2
varies
6.5 tok/s
4.5–8.4
lmstudio-community/DeepSeek-Coder-V2-Lite-Instruct-GGUF
9 runs · 5 configs
★ bestllama.cpp8k ctxthinking default
×2 runs
49.843.5–56.2
varies
17.9 tok/s
17.9–17.9
bartowski/Qwen2.5-7B-Instruct-GGUF
10 runs · 6 configs
★ bestllama.cpp66k ctxthinking default
×2 runs
48.444.2–52.5
varies
10.4 tok/s
7.8–13.0
qwen2.5-coder
3 runs · 3 configs
Compare with another machine
Run history
Oct 12, 2026
20:24Llama-3.2-8X3B-MOE-Dark-Champion-Instruct-uncensored-abliterated-18.4B-GGUFllama.cpp3.5 tok/s40.720:06Llama-3.2-8X3B-MOE-Dark-Champion-Instruct-uncensored-abliterated-18.4B-GGUFllama.cpp4.2 tok/s38.619:51Llama-3.2-8X3B-MOE-Dark-Champion-Instruct-uncensored-abliterated-18.4B-GGUFllama.cpp5.0 tok/s33.619:35Llama-3.2-8X3B-MOE-Dark-Champion-Instruct-uncensored-abliterated-18.4B-GGUFllama.cpp4.4 tok/s32.919:13Llama-3.2-8X3B-MOE-Dark-Champion-Instruct-uncensored-abliterated-18.4B-GGUFllama.cpp8.9 tok/s25.915:22DeepSeek-Coder-V2-Lite-Instruct-GGUFllama.cpp1.5 tok/s53.814:37DeepSeek-Coder-V2-Lite-Instruct-GGUFllama.cpp13.7 tok/s53.514:30DeepSeek-Coder-V2-Lite-Instruct-GGUFllama.cpp15.2 tok/s48.314:25DeepSeek-Coder-V2-Lite-Instruct-GGUFllama.cpp13.7 tok/s42.514:18DeepSeek-Coder-V2-Lite-Instruct-GGUFllama.cpp17.9 tok/s56.214:12Qwen2.5-Coder-14B-Instruct-GGUFllama.cpp5.3 tok/s49.413:55Qwen2.5-Coder-14B-Instruct-GGUFllama.cpp4.5 tok/s61.213:37Qwen2.5-Coder-14B-Instruct-GGUFllama.cpp8.2 tok/s51.013:25Qwen2.5-Coder-14B-Instruct-GGUFllama.cpp6.5 tok/s44.213:12Qwen2.5-Coder-14B-Instruct-GGUFllama.cpp6.0 tok/s43.112:42gemma-4-12B-agentic-fable5-composer2.5-v2-3.5x-tau2-GGUFllama.cpp5.0 tok/s65.212:24gemma-4-12B-agentic-fable5-composer2.5-v2-3.5x-tau2-GGUFllama.cpp6.7 tok/s52.212:11gemma-4-12B-agentic-fable5-composer2.5-v2-3.5x-tau2-GGUFllama.cpp5.7 tok/s65.311:55gemma-4-12B-agentic-fable5-composer2.5-v2-3.5x-tau2-GGUFllama.cpp6.7 tok/s65.011:42gemma-4-12B-agentic-fable5-composer2.5-v2-3.5x-tau2-GGUFllama.cpp7.4 tok/s45.210:13Qwen2.5-Coder-7B-Instruct-GGUFllama.cpp7.1 tok/s35.110:04Qwen2.5-Coder-7B-Instruct-GGUFllama.cpp10.2 tok/s36.209:57Qwen2.5-Coder-7B-Instruct-GGUFllama.cpp10.1 tok/s40.309:50Qwen2.5-Coder-7B-Instruct-GGUFllama.cpp10.1 tok/s44.909:41Qwen2.5-Coder-7B-Instruct-GGUFllama.cpp5.7 tok/s45.309:31Qwen3-Coder-30B-A3B-Instruct-GGUFllama.cpp10.1 tok/s54.309:19Qwen3-Coder-30B-A3B-Instruct-GGUFllama.cpp12.6 tok/s63.509:08Qwen3-Coder-30B-A3B-Instruct-GGUFllama.cpp13.8 tok/s61.909:00Qwen3-Coder-30B-A3B-Instruct-GGUFllama.cpp14.2 tok/s62.608:50Qwen3-Coder-30B-A3B-Instruct-GGUFllama.cpp15.9 tok/s60.607:16Meta-Llama-3.1-8B-Instruct-abliterated-GGUFllama.cpp6.9 tok/s20.605:37Meta-Llama-3.1-8B-Instruct-abliterated-GGUFllama.cpp9.8 tok/s14.305:23Qwen3-14B-abliterated-GGUFllama.cpp4.3 tok/s46.204:37Qwen3-14B-abliterated-GGUFllama.cpp6.7 tok/s65.404:21Qwen3-14B-abliterated-GGUFllama.cpp6.5 tok/s47.803:49Qwen3-14B-abliterated-GGUFllama.cpp6.1 tok/s64.703:21Qwen3-14B-abliterated-GGUFllama.cpp8.2 tok/s55.503:00Qwen2.5-7B-Instruct-GGUFllama.cpp11.4 tok/s39.702:53Qwen2.5-7B-Instruct-GGUFllama.cpp13.0 tok/s52.502:46Qwen2.5-7B-Instruct-GGUFllama.cpp13.0 tok/s43.802:40Qwen2.5-7B-Instruct-GGUFllama.cpp13.0 tok/s47.002:33Qwen2.5-7B-Instruct-GGUFllama.cpp13.0 tok/s46.0Oct 10, 2026
22:34DeepSeek-Coder-V2-Lite-Instruct-GGUFllama.cpp18.3 tok/s45.822:29DeepSeek-Coder-V2-Lite-Instruct-GGUFllama.cpp17.9 tok/s39.422:23DeepSeek-Coder-V2-Lite-Instruct-GGUFllama.cpp17.6 tok/s34.922:18DeepSeek-Coder-V2-Lite-Instruct-GGUFllama.cpp17.9 tok/s43.522:13Qwen2.5-Coder-14B-Instruct-GGUFllama.cpp8.4 tok/s46.722:03Qwen2.5-Coder-14B-Instruct-GGUFllama.cpp8.4 tok/s56.421:52Qwen2.5-Coder-14B-Instruct-GGUFllama.cpp8.4 tok/s54.621:43Qwen2.5-Coder-14B-Instruct-GGUFllama.cpp8.3 tok/s58.621:33gemma-4-12B-agentic-fable5-composer2.5-v2-3.5x-tau2-GGUFllama.cpp7.4 tok/s66.221:20gemma-4-12B-agentic-fable5-composer2.5-v2-3.5x-tau2-GGUFllama.cpp7.4 tok/s54.521:08gemma-4-12B-agentic-fable5-composer2.5-v2-3.5x-tau2-GGUFllama.cpp7.4 tok/s59.020:55gemma-4-12B-agentic-fable5-composer2.5-v2-3.5x-tau2-GGUFllama.cpp7.4 tok/s53.320:43Qwen2.5-Coder-7B-Instruct-GGUFllama.cpp10.2 tok/s41.620:04Qwen2.5-Coder-7B-Instruct-GGUFllama.cpp7.7 tok/s25.019:22Qwen2.5-Coder-7B-Instruct-GGUFllama.cpp9.9 tok/s29.319:12Qwen3-Coder-30B-A3B-Instruct-GGUFllama.cpp22.8 tok/s64.019:06Qwen3-Coder-30B-A3B-Instruct-GGUFllama.cpp23.6 tok/s69.418:58Qwen3-Coder-30B-A3B-Instruct-GGUFllama.cpp23.2 tok/s64.218:48Qwen3-Coder-30B-A3B-Instruct-GGUFllama.cpp22.7 tok/s66.618:40Meta-Llama-3.1-8B-Instruct-abliterated-GGUFllama.cpp12.4 tok/s18.518:06Meta-Llama-3.1-8B-Instruct-abliterated-GGUFllama.cpp10.5 tok/s9.317:10Meta-Llama-3.1-8B-Instruct-abliterated-GGUFllama.cpp11.9 tok/s10.816:58Qwen3-14B-abliterated-GGUFllama.cpp8.1 tok/s49.315:59Qwen3-14B-abliterated-GGUFllama.cpp7.7 tok/s51.415:14Qwen3-14B-abliterated-GGUFllama.cpp8.1 tok/s55.514:56Qwen2.5-7B-Instruct-GGUFllama.cpp7.8 tok/s44.214:29Qwen2.5-7B-Instruct-GGUFllama.cpp9.9 tok/s47.614:20Qwen2.5-7B-Instruct-GGUFllama.cpp9.9 tok/s43.214:12Qwen2.5-7B-Instruct-GGUFllama.cpp10.0 tok/s38.014:03Qwen2.5-7B-Instruct-GGUFllama.cpp10.0 tok/s43.5Oct 8, 2026
11:08qwen3.5:latestOllama15.0 tok/s61.510:44qwen3.5:latestOllama15.0 tok/s57.510:21qwen3.5:latestOllama15.1 tok/s64.010:02qwen3.5:latestOllama14.9 tok/s57.809:40qwen2.5-coder:latestOllama18.4 tok/s39.209:36qwen2.5-coder:latestOllama18.5 tok/s42.609:32qwen2.5-coder:latestOllama18.4 tok/s26.809:23qwen3-coder:latestOllama13.2 tok/s67.409:14qwen3-coder:latestOllama14.0 tok/s63.309:06qwen3-coder:latestOllama15.0 tok/s58.201:59gemma3:latestOllama20.8 tok/s54.101:54gemma3:latestOllama21.9 tok/s53.001:48gemma3:latestOllama21.6 tok/s49.601:23gemma3:latestOllama25.0 tok/s55.5