Machine

RTX 5090 · Ryzen 9 7900X #a7375cec

20 runs of 15 models since Aug 25, 2026. Last run Nov 16, 2026.

ShareRedditX
20
runs
15
models
17
configs
3
configs run 2+ times

Hardware

GPU
NVIDIA GeForce RTX 5090 32 GB
CPU
AMD Ryzen 9 7900X 12-Core Processor 12 cores
RAM
62 GB
OS
Fedora Linux 45 (KDE Plasma Desktop Edition)
Tools
vLLM 14 · llama.cpp 3 · Strata 3
Client
0.4.99
All NVIDIA GeForce RTX 5090 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 read4050607080901002050100200tok/s (log scale) →quality →

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.

qwen3.8-27b
1 run · 1 config
★ bestvLLM64k ctxthinking default
89.1
1 rated run
155 tok/s
Qwen3.8-27B-QUASAR-NVFP4
1 run · 1 config
★ bestvLLM64k ctxthinking default
88.3
1 rated run
79.9 tok/s
unsloth/Qwen3.8-Flash-Next-GGUF
2 runs · 1 config
★ bestvLLM64k ctxthinking default
×2 runs
87.787.2–88.1
steady
39.1 tok/s
34.9–43.4
AtomicChat/Qwen3.8-Flash-Next-GGUF
1 run · 1 config
★ bestvLLM64k ctxthinking default
86.9
1 rated run
39.7 tok/s
esatapedico/Qwen3.8-27B-NVFP4-MTP-GGUF
1 run · 1 config
★ bestvLLM64k ctxthinking on
86.9
1 rated run
134 tok/s
swift-1.5-iq3_xxs
1 run · 1 config
★ bestStrata256k ctxthinking oneffort xhigh
85.9
1 rated run
117 tok/s
Qwen3.8-27B QUASAR NVFP4
1 run · 1 config
★ bestllama.cpp256k ctxthinking defaulteffort xhigh
85.1
1 rated run
247 tok/s
SC117/Ornith-1.5-35B-A3B-Heretic-MTP-APEX-GGUF
1 run · 1 config
★ bestvLLM64k ctxthinking default
84.9
1 rated run
222 tok/s
qwen3.6-35b-a3b
1 run · 1 config
★ bestvLLM64k ctxthinking default
84.7
1 rated run
437 tok/s
qwen3.8-flash-next-iq3_s
2 runs · 2 configs
★ bestStrata256k ctxthinking defaulteffort xhigh
84.5
1 rated run
122 tok/s

Compare with another machine

Run history