Full Benchmark (16 Combinations)
Accuracy vs. Speed from the first benchmark run.
Synchronous Combinations (3-Test Average)
X-axis: Speed (left=better) | Y-axis: Accuracy (top=better) | Bubble Size: VRAM (small=better).
Results for Qwen3-30B on LongBench-v2
View Project on GitHubAccuracy vs. Speed from the first benchmark run.
X-axis: Speed (left=better) | Y-axis: Accuracy (top=better) | Bubble Size: VRAM (small=better).