Machine
RTX 2080 Ti + RTX A2000 12gb · Ryzen 7 5800X #0d84416a
39 runs of 23 models since Oct 11, 2026. Last run Oct 20, 2026.
39
runs
23
models
29
configs
7
configs run 2+ times
Hardware
- GPU
- NVIDIA GeForce RTX 2080 Ti + NVIDIA RTX A2000 12GB 23 GB
- CPU
- AMD Ryzen 7 5800X 8-Core Processor 8 cores
- RAM
- 20 GB
- OS
- Debian GNU/Linux 13 (trixie)
- Tools
- llama.cpp 39
- Client
- 0.4.96
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, 23 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
nex-n2.5-mini
3 runs · 1 config
★ bestllama.cpp64k ctxthinking off
×3 runs
84.977.8–85.5
steady
65.9 tok/s
65.8–66.5
underdog-saluki
1 run · 1 config
opus-v2
1 run · 1 config
qwen38-27b-turbo
3 runs · 2 configs
★ bestllama.cpp8k ctxthinking default
×2 runs
77.877.7–78.0
steady
29.7 tok/s
29.5–29.8
gpt-oss-20b
1 run · 1 config
qwen35-27b-opus
1 run · 1 config
qwen3.8-35b-miniplus
2 runs · 2 configs
lfm2.5-24b-a2b
1 run · 1 config
ornith-heretic-mini
1 run · 1 config
ornith15-heretic-mini
3 runs · 1 config
★ bestllama.cpp64k ctxthinking off
×3 runs
74.774.5–77.7
steady
72.7 tok/s
72.2–73.6
Compare with another machine
Run history
Oct 20, 2026
22:29qwen35-27b-opusllama.cpp17.5 tok/s76.922:05opus-v1-officialllama.cpp77.9 tok/s70.322:01opus-v2llama.cpp73.6 tok/s79.321:52qwen38-27b-iq2llama.cpp27.0 tok/s67.909:21ternary-bonsai-ptq1-samplerllama.cpp56.2 tok/s65.5Oct 19, 2026
16:45ternary-bonsai-ptq1llama.cpp58.3 tok/s74.7Oct 16, 2026
20:52nemotron-lightning-30b-a3bllama.cpp86.7 tok/s74.020:44gemma4-26b-uncensored-mtpllama.cpp59.9 tok/s70.120:42gemma4-26b-a4b-qatllama.cpp64.8 tok/s68.120:29minicpm-v47-35b-a3bllama.cpp71.2 tok/s73.520:27gpt-oss-20bllama.cpp80.6 tok/s77.320:24qwen36-35b-a3b-mtpllama.cpp83.2 tok/s65.120:21ornith15-heretic-minillama.cpp72.7 tok/s74.720:18gemma4-26b-opus-distillllama.cpp62.6 tok/s72.820:15tiel-coder-35bllama.cpp75.7 tok/s74.620:14nex-n2.5-minillama.cpp66.5 tok/s84.9Oct 15, 2026
21:13gemma4-26b-uncensored-mtpllama.cpp59.5 tok/s69.621:11ornith15-heretic-minillama.cpp73.6 tok/s77.721:08nex-n2.5-minillama.cpp65.8 tok/s85.520:58gemma4-26b-a4b-qatllama.cpp65.5 tok/s67.520:56gemma4-26b-opus-distillllama.cpp63.1 tok/s70.520:54tiel-coder-35bllama.cpp78.6 tok/s70.619:29gemma4-26b-uncensored-mtpllama.cpp58.7 tok/s71.519:27qwen38-27b-turbollama.cpp29.3 tok/s71.119:23qwen3.8-35b-miniplusllama.cpp77.6 tok/s72.019:15ornith15-heretic-minillama.cpp72.2 tok/s74.519:10nex-n2.5-minillama.cpp65.9 tok/s77.818:56lfm2.5-24b-a2bllama.cpp128 tok/s75.018:44gemma4-26b-opus-distillllama.cpp63.5 tok/s65.518:40tiel-coder-35bllama.cpp73.5 tok/s83.418:37underdog-salukillama.cpp24.9 tok/s80.418:32qwen38-coder390llama.cpp16.4 tok/s67.918:22gemma4-31b-qatllama.cpp16.7 tok/s72.718:16ornith-heretic-minillama.cpp71.3 tok/s75.0