Machine

Gtx 1050 Ti + RX 7800 Xt · Core i7-7700K #c6793251

76 runs of 26 models since Oct 4, 2026. Last run Nov 18, 2026.

ShareRedditX
76
runs
26
models
59
configs
11
configs run 2+ times

Hardware

GPU
NVIDIA GeForce GTX 1050 Ti + AMD Radeon RX 7800 XT 20 GB
CPU
Intel(R) Core(TM) i7-7700K CPU @ 4.20GHz 4 cores
RAM
31 GB
OS
CachyOS
Tools
llama.cpp 44 · llama.cpp k8 v4 -1 budget 9 · OpenAI-compatible 4 · llama.cpp k8 v4 4 · llama.cpp k8 vTurbo4 1 · llama.cpp k4 v4 1 · llama.cpp k5 v5 16384 budget 1 · llama.cpp k8 v4 16384 budget 1 · llama.cpp k8 v4 12288 budget 1 · llama.cpp k4 v4 8192 budget 1 · llama.cpp k4 v4 4096 budget 1 · llama.cpp k4 v4 2048 budget 1 · llama.cpp k5 v5 1 · llama.cpp k8 v8 -1 budget 1 · llama.cpp q5 q5 -1 budget 1 · llama.cpp q8 q4 -1 budget 1 · llama.cpp q8 q8 -1 budget 1 · llama.cpp q8 q8 1 · Strata 1
Client
0.4.99
All NVIDIA GeForce GTX 1050 Ti + AMD Radeon RX 7800 XT benchmarks

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
below 20 tok/s: slower than you read4050607080901002050100tok/s (log scale) →quality →

Medians of each model's best config, 26 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.

byteshape/Qwen3.8-27B-GGUF
8 runs · 6 configs
★ bestIQ4_XS-ASCII-Condensedllama.cpp64k ctxthinking oneffort xhigh
×2 runs
89.789.0–90.4
steady
41.6 tok/s
41.3–41.9
unsloth/Qwen3.8-27B-UD-Q2_K_XL
6 runs · 2 configs
★ bestQ2_Kllama.cpp64k ctxthinking oneffort xhigh
×5 runs
89.085.9–90.4
steady
45.4 tok/s
44.2–46.0
Qwen3.8-27B-Turbo-Brilliance-50X-RI-GSQ-RCO-IQ3_XXS-mtp-ASCII-Condensed
1 run · 1 config
★ bestllama.cpp k8 v4 -1 budget64k ctxthinking on
88.9
1 rated run
35.9 tok/s
Qwen3.8-27B-Turbo-Brilliance-50X-RI-GSQ-RCO-IQ3_XXS-mtp
3 runs · 3 configs
★ bestOpenAI-compatible64k ctxthinking default
88.8
1 rated run
34.6 tok/s
Qwen3.8-27B-GSQ-RCO-BuffedMod-IQ3_S_XL-mtp-Q6K
2 runs · 2 configs
★ bestllama.cpp k8 v464k ctxthinking default
87.8
1 rated run
37.9 tok/s
Swift-1.5-IQ3_XSS
1 run · 1 config
★ bestOpenAI-compatible125k ctxthinking on
87.6
1 rated run
48.9 tok/s
Swift-1.5-Qwen3.8-27B-GSQ-RCO-IQ3_XXS-mtp
14 runs · 12 configs
★ bestllama.cpp128k ctxthinking defaulteffort xhigh
×3 runs
86.685.6–87.3
steady
51.5 tok/s
47.4–52.2
Qwen3.8-27B-GSQ-RCO-IQ3_S-mtp-ASCII-Condensed
7 runs · 6 configs
★ bestllama.cpp k8 v4 -1 budget125k ctxthinking on
×2 runs
86.683.4–89.8
steady
49.4 tok/s
49.0–49.8
Qwen3.8-27B-GSQ-RCO-IQ3_S-recovered
4 runs · 2 configs
★ bestllama.cpp64k ctxthinking defaulteffort xhigh
×3 runs
86.283.8–86.7
steady
34.1 tok/s
33.8–40.7
Swift-1.5-Qwen3.8-27B-GSQ-RCO-IQ3_S-mtp
5 runs · 4 configs
★ bestllama.cpp64k ctxthinking defaulteffort xhigh
×2 runs
85.683.1–88.1
steady
36.2 tok/s
34.6–37.7

Compare with another machine

Run history

Show all 76 runs ▾