Qwen
Benchmarks
TaskDatasetSOTA resultTrendResults
Qwen-3-0.6B
14.54Input Success Rate (%)
9
Qwen VL-235B 3
2.32ASR
9
Qwen2.5-VL-32B
10.8ASR
9
Qwen2.5-VL-7B
98ASR
9
Qwen2.5-7B
1,480.2Throughput (tokens/s)
9
Qwen3-8B
26.74Throughput (tokens/s)
9
Qwen2.5-VL-7B
89.2Accuracy
9
Qwen VL 2.5
0.258CLIP Similarity (RN-50)
9
Qwen1.5-MoE
100WSR
8
Qwen3-4B S=4096 (base variant)
214Per-Step Latency (ms)
8
Qwen3-30B-A3B BF16 128 experts top-8
26.79Throughput (tokens/s)
8
Qwen 27B 3.6
89ASR
8
Qwen-7B-Instruct SFT Variant 2.5
0FSR
8
Qwen 7B 2.5 (Anchor)
0FSR
8
Qwen 3B 2.5
24.6∆PPL (%)
8
Qwen VL 32B 3
11.7Attack Success Rate (ASR)
8
Qwen3-VL-8B
18.3ASR
8
Qwen3-VL-32B
77.6ASR
8
Qwen3 VL 8B
80.6ASR
8
Qwen 14B 2.5
1AUC
7
Qwen 7B 2.5
100AUC
7
Qwen2.5-7B (test)
37.29Throughput
7
Qwen models
0.063ECE2
7
Qwen 7B 2.5
2.7Model Size (GB)
7
Qwen2.5-7B 256K context (train)
1,301.6Throughput (tokens/sec)
7