Machine

RTX 3080 + RTX 3080 Ti · Ryzen 7 5800X #e42f80d7

35 runs of 21 models since Aug 19, 2026. Last run Nov 7, 2026.

ShareRedditX
35
runs
21
models
30
configs
3
configs run 2+ times

Hardware

GPU
NVIDIA GeForce RTX 3080 + NVIDIA GeForce RTX 3080 Ti 32 GB
CPU
AMD Ryzen 7 5800X 8-Core Processor 8 cores
RAM
62 GB
OS
Arch Linux
Tools
llama.cpp 28 · SGLang 5 · vLLM 2
Client
0.4.99
All NVIDIA GeForce RTX 3080 + NVIDIA GeForce RTX 3080 Ti 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 read1030507090102050100200tok/s (log scale) →quality →outliers sit at the edges

Medians of each model's best config, 21 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.

cyankiwi/Qwen3.8-27B-AWQ-INT4
2 runs · 2 configs
★ bestINT4SGLang256k ctxthinking on
89.6
1 rated run
37.2 tok/s
unsloth/Qwen3.8-27B-GGUF
1 run · 1 config
★ bestUD-Q6_K_XLllama.cpp64k ctxthinking on
89.6
1 rated run
47.1 tok/s
unsloth/Muse-Glimmer-30B-GGUF
3 runs · 3 configs
★ bestUD-Q6_K_XLllama.cpp64k ctxthinking default
87.6
1 rated run
52.3 tok/s
Qwen3.8-27B-SC_6.00bpw_H6
1 run · 1 config
★ bestllama.cpp64k ctxthinking on
86.4
1 rated run
41.8 tok/s
Jackrong/Qwopus3.8-27B-Coder-Flash-GGUF
1 run · 1 config
★ bestQ6_Kllama.cpp64k ctxthinking oneffort xhigh
86.1
1 rated run
62.3 tok/s
unsloth/Qwen3.8-Flash-Next-GGUF
1 run · 1 config
★ bestUD-Q4_K_XLllama.cpp64k ctxthinking on
85.7
1 rated run
21.5 tok/s
unsloth/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF
5 runs · 2 configs
★ bestUD-Q4_K_XLllama.cpp64k ctxthinking on
×4 runs
79.176.2–85.6
varies
179 tok/s
159–184
cyankiwi/gemma-4-31B-it-AWQ-4bit
4 runs · 4 configs
★ bestSGLang256k ctxthinking on
78.8
1 rated run
32.0 tok/s
unsloth/NVIDIA-Nemotron-3.5-Bolt-74B-A9B-GGUF
1 run · 1 config
★ bestIQ2_XXSllama.cpp64k ctxthinking on
78.7
1 rated run
79.4 tok/s
./gemma-4-31B-it-AWQ4-styletune-head
1 run · 1 config
★ bestSGLang256k ctxthinking on
78.6
1 rated run
32.1 tok/s

Compare with another machine

Run history

Show all 35 runs ▾