Community performance summary

AMD Radeon RX 6600 XT

4 local models measured on this GPU configuration. Compare quantizations, configurations and the full observed performance range.

9 signed runs · 0 verified · 6 testers
Median prompt166.97tok/s · 164.13 to 1971.4
Median generation19.51tok/s · 13.97 to 89.79
Configurations4Q4_K_M · Q8_0 · UD-Q4_K_M
Last observedJul 20RDNA 2

What the results show

One clear view.
Every recorded run.

Measurements from comparable hardware and models are presented together. Medians show typical performance, while every signed run remains available below for inspection and sharing.

Models on this GPU4 groups
Qwen3 VL 2B1 runGemma 3 4B1 runQwen3.6 35B-A3B6 runsGemma 4 26B-A4B1 run
macOS

macOS 15.2 Sequoia · macOS 15.7.7 Sequoia

Recorded evidence

Inspect every run.

Open any result to review its exact hardware, model configuration, measurements and submission details.

1 to 8 of 9 configurations

Verified benchmark Community submission
Model / quantHardwareArchitecturePromptGenerationSource
Qwen3 VL 2BQ8_0 · Dense · Signed ToshLLM app · amd-gpu · Metal · comparison
RDNA 2
1971.4tok/s
89.79tok/s
ardentwheelJul 20, 2026
Gemma 3 4BQ4_K_M · Dense · Signed ToshLLM app · amd-gpu · Metal · comparison
RDNA 2
899.15tok/s
62.76tok/s
ArdentwheelJul 20, 2026
Qwen3.6 35B-A3BUD-Q4_K_S · MoE · Signed ToshLLM app · ncmoe 32 · amd-gpu · Metal · comparison
RDNA 2
167.83tok/s
20.06tok/s
test1-Henbury-DIsableJul 20, 2026
Qwen3.6 35B-A3BUD-Q4_K_S · MoE · Signed ToshLLM app · ncmoe 32 · amd-gpu · Metal · comparison
RDNA 2
164.3tok/s
19.86tok/s
ardentwheelJul 19, 2026
Qwen3.6 35B-A3BUD-Q4_K_S · MoE · Signed ToshLLM app · ncmoe 32 · amd-gpu · Metal · comparison
RDNA 2
166.97tok/s
19.51tok/s
test0-Henbury-DIsableJul 20, 2026
Gemma 4 26B-A4BUD-Q4_K_M · MoE · Signed ToshLLM app · ncmoe 22 · amd-gpu · Metal · comparison
RDNA 2
231.85tok/s
17.29tok/s
ArdentwheelJul 20, 2026
Qwen3.6 35B-A3BUD-Q4_K_S · MoE · Signed ToshLLM app · ncmoe 32 · amd-gpu · Metal · comparison
RDNA 2
164.7tok/s
14.07tok/s
ardentwheelJul 18, 2026
Qwen3.6 35B-A3BUD-Q4_K_S · MoE · Signed ToshLLM app · ncmoe 32 · amd-gpu · Metal · comparison
RDNA 2
164.44tok/s
14.03tok/s
test1-Henbury-EnableJul 20, 2026
Qwen3.6 35B-A3BUD-Q4_K_S · MoE · Signed ToshLLM app · ncmoe 32 · amd-gpu · Metal · comparison
RDNA 2
164.13tok/s
13.97tok/s
test0-Henbury-EnableJul 20, 2026