Machine
RX 7900 XTX · Ryzen 7 5700X3D #75174f68
58 runs of 14 models since Aug 28, 2026. Last run Oct 29, 2026.
58
runs
14
models
22
configs
10
configs run 2+ times
Hardware
- GPU
- AMD Radeon RX 7900 XTX 24 GB
- CPU
- AMD Ryzen 7 5700X3D 8-Core Processor 8 cores
- RAM
- 46 GB
- OS
- Nobara Linux 44 (KDE Plasma Desktop Edition)
- Tools
- llama.cpp 52 · Strata 6
- Client
- 0.4.87
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, 14 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
swift-1.5-iq2_xs
5 runs · 5 configs
Qwen3.8-27B-Q4_K_XL-rdna-boosts
1 run · 1 config
Qwen3.8-27B-IQ3_XXS
1 run · 1 config
Tiel-Coder-35B-A3B-Q4_K_S
5 runs · 2 configs
★ bestQ4_K_Sllama.cpp64k ctxthinking on
×4 runs
85.083.8–87.5
steady
164 tok/s
160–167
Qwen3.6-35B-A3B-Q4_K_XL
2 runs · 1 config
★ bestQ4_Kllama.cpp64k ctxthinking on
×2 runs
84.083.9–84.1
steady
157 tok/s
155–159
Qwen3.8-27B-Q4_K_XL
15 runs · 2 configs
★ bestQ4_Kllama.cpp64k ctxthinking on
×13 runs
82.974.2–87.1
varies
73.8 tok/s
54.1–77.0
Qwen3.8-27B-Q5-XYZ-v2
1 run · 1 config
Qwen3.8-27B-ROCmFP4-FAST
1 run · 1 config
Ornith-1.5-35B-A3B-IQ4_XS
3 runs · 1 config
★ bestllama.cpp64k ctxthinking on
×3 runs
78.978.4–84.8
steady
140 tok/s
139–140
Qwen3.6-35B-A3B-IQ4_NL
17 runs · 3 configs
★ bestllama.cpp8k ctxthinking default
×3 runs
76.973.4–83.5
varies
161 tok/s
161–170
Compare with another machine
Run history
Oct 29, 2026
11:49LFM3-1.7B-Q8_0llama.cpp267 tok/s69.9Oct 11, 2026
08:45swift-1.5-iq2_xsStrata76.4 tok/s87.208:31Qwen3.8-27B-ROCmFP4-FASTllama.cpp86.2 tok/s79.101:56swift-1.5-iq2_xsStrata60.2 tok/s85.5Oct 5, 2026
07:25swift-1.5-iq2_xsStrata86.1 tok/s85.707:17swift-1.5-iq2_xsStrata91.4 tok/s48.507:09swift-1.5-iq2_xsStrata86.7 tok/s81.7Sep 19, 2026
22:55Tiel-Coder-35B-A3B-Q4_K_Sllama.cpp158 tok/s84.222:51Tiel-Coder-35B-A3B-Q4_K_Sllama.cpp160 tok/s87.506:15Qwen3.8-9B-Q8_0×2llama.cpp126 tok/s68.3Show all 58 runs ▾Show fewer ▴
Sep 5, 2026
14:06gemma-4-26b-a4b-q4kxl×3llama.cpp171 tok/s69.514:00Tiel-Coder-35B-A3B-Q4_K_S×3llama.cpp166 tok/s84.913:51Ornith-1.5-35B-A3B-IQ4_XS×3llama.cpp140 tok/s78.913:38Qwen3.6-35B-A3B-IQ4_NL×3llama.cpp163 tok/s77.113:32Qwen3.8-27B-Q4_K_XL×3llama.cpp71.3 tok/s83.813:08Qwen3.6-35B-A3B-IQ4_NL×2llama.cpp160 tok/s74.6Sep 1, 2026
13:27Qwen3.8-27B-Q4_K_XLllama.cpp72.2 tok/s67.811:49Qwen3.8-27B-Q4_K_XL-rdna-boostsllama.cpp57.2 tok/s86.508:16Qwen3.6-35B-A3B-IQ4_NL×3llama.cpp161 tok/s74.207:38Qwen3.6-35B-A3B-IQ4_NL×2llama.cpp161 tok/s80.207:10Qwen3.6-35B-A3B-IQ4_NLllama.cpp162 tok/s76.7Aug 30, 2026
09:05Qwen3.6-35B-A3B-Q4_K_XL×2llama.cpp157 tok/s84.008:57Qwen3.6-35B-A3B-IQ4_NL×4llama.cpp174 tok/s75.102:58Qwen3.6-35B-A3B-IQ4_NLllama.cpp170 tok/s73.402:44Qwen3.6-35B-A3B-IQ4_NLllama.cpp165 tok/s73.601:58Qwen3.8-27B-Q4_K_XL×2llama.cpp75.4 tok/s81.0