Machine
RTX 4070 · Core i5-13400F #024030e9
32 runs of 15 models since Nov 13, 2026. Last run Nov 20, 2026.
32
runs
15
models
22
configs
7
configs run 2+ times
Hardware
- GPU
- NVIDIA GeForce RTX 4070 12 GB
- CPU
- 13th Gen Intel(R) Core(TM) i5-13400F 10 cores
- RAM
- 63 GB
- OS
- Windows 11 Pro 26H2
- Tools
- llama.cpp 31 · Strata 1
- Client
- 0.4.99
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, 15 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
qwen3.8-flash-next-iq3_xxs
3 runs · 3 configs
ornith-1.5-35b-a3b-apex-mtp-quality
3 runs · 1 config
★ bestllama.cpp64k ctxthinking default
×3 runs
86.385.3–88.0
steady
104 tok/s
102–104
cyber-tiel-coder-35b-a3b-mtp-ud-q5_k_xl
5 runs · 2 configs
★ bestQ5_Kllama.cpp64k ctxthinking default
×4 runs
85.683.3–87.9
steady
104 tok/s
90.6–112
ornith-1.5-35b-a3b-apex-mtp-i-quality
2 runs · 1 config
★ bestllama.cpp64k ctxthinking default
×2 runs
85.484.8–86.0
steady
105 tok/s
104–106
default
4 runs · 3 configs
★ bestllama.cpp64k ctxthinking oneffort medium
×2 runs
85.182.9–87.4
steady
68.9 tok/s
66.7–71.0
cyber-tiel-coder-35b-a3b-mtp-ud-q6_k_xl
2 runs · 1 config
★ bestQ6_Kllama.cpp64k ctxthinking default
×2 runs
84.082.8–85.2
steady
92.4 tok/s
91.7–93.0
cyber-tiel-coder-35b-a3b-mtp-ud-q4_k_xl
1 run · 1 config
ornith-1.5-35b-a3b-crack-q4_k_m
2 runs · 2 configs
ornith-1.5-35b-a3b-apex-i-miniplus-v2.1
1 run · 1 config
qwen3.6-35b-a3b-uncensored-hauhaucs-aggressive-q4_k_m
1 run · 1 config
Compare with another machine
Run history
Nov 20, 2026
15:53defaultllama.cpp66.7 tok/s82.915:48defaultllama.cpp69.8 tok/s78.115:41defaultllama.cpp71.0 tok/s87.415:35defaultllama.cpp67.4 tok/s83.1Nov 16, 2026
10:20sage-a1.5-q4_k_mllama.cpp120 tok/s75.4Nov 15, 2026
10:06ornith-1.5-35b-a3b-apex-mtp-qualityllama.cpp102 tok/s85.309:58cyber-tiel-coder-35b-a3b-mtp-ud-q5_k_xl×2llama.cpp108 tok/s83.909:37qwen3.8-35b-a3b-distill-miniplus-v2.1-abliterated×2llama.cpp139 tok/s77.409:09qwen3.8-flash-next-iq3_xxsllama.cpp52.3 tok/s81.509:01qwen3.8-flash-next-iq3_xxsllama.cpp48.2 tok/s87.700:22qwen3.8-flash-next-iq3_sllama.cpp29.7 tok/s76.0Nov 14, 2026
15:32ornith-1.5-35b-a3b-apex-mtp-qualityllama.cpp104 tok/s86.315:22ornith-1.5-35b-a3b-apex-mtp-i-quality×2llama.cpp105 tok/s85.414:52ornith-1.5-35b-a3b-apex-mtp-qualityllama.cpp104 tok/s88.014:34ornith-1.5-35b-a3b-apex-i-miniplus-v2.1llama.cpp138 tok/s80.713:07cyber-tiel-coder-35b-a3b-mtp-ud-q6_k_xl×2llama.cpp92.4 tok/s84.012:08huihui-qwen3.6-35b-a3b-abliterated-q5_kllama.cpp130 tok/s78.212:02cyber-tiel-coder-35b-a3b-mtp-ud-q5_k_xlllama.cpp104 tok/s87.9Show all 32 runs ▾Show fewer ▴
Nov 13, 2026
21:54qwen3.6-35b-a3b-uncensored-hauhaucs-aggressive-q4_k_mllama.cpp124 tok/s79.621:48ornith-1.5-35b-a3b-crack-q4_k_mllama.cpp102 tok/s81.221:40ornith-1.5-35b-a3b-crack-q4_k_mllama.cpp107 tok/s65.321:31minicpm-v-4.7-35b-a3b-uncensored-ad-q5_k-q4_kllama.cpp104 tok/s77.421:16huihui-qwen3.6-35b-a3b-abliterated-q5_kllama.cpp135 tok/s81.321:11huihui-qwen3.6-35b-a3b-abliterated-q5_kllama.cpp138 tok/s66.821:03cyber-tiel-coder-35b-a3b-mtp-ud-q4_k_xlllama.cpp122 tok/s82.420:54cyber-tiel-coder-35b-a3b-mtp-ud-q5_k_xlllama.cpp104 tok/s86.720:46cyber-tiel-coder-35b-a3b-mtp-ud-q5_k_xlllama.cpp90.6 tok/s86.8