Qwen
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
Qwen max 3
96.3Attack Success Rate (ASR)
4
Qwen-2.5 series
0.971SeedPrint
4
Qwen 1.5B Instruct 2.5
155.3Throughput (tok/s)
4
Qwen3-VL-8B
29.5Peak VRAM (RTX)
4
Qwen VL-32B 2.5
98.4Peak VRAM (H200)
4
Qwen2.5-7B
3.22PPL
4
Qwen2.5-7B 256K context length
26.3Decode Latency (ms/token)
4
Qwen2.5-7B 128K context length
18.4Decode Latency (ms/token)
4
Qwen2.5-7B 64K context length
13.6Decode Latency (ms/token)
4
Qwen2.5-7B 32K context length
10.3Decoding Latency (ms/token)
4
Qwen2.5-7B 16K context length
8.4Decode Latency (ms/token)
4
Qwen2.5-7B 4K context length
6.2Decode Latency (ms/token)
4
Qwen2.5 0.5B
139.98Latency (ms)
4
Qwen Subliminal
80.07Explanatory Variance
4
Qwen 2.5 14B 2048 input + 8192 generation tokens
287.7End-to-end Latency (s)
4
Qwen 14B (2048 input + 2048 generation tokens) 2.5
70.3End-to-end Latency (s)
4
Qwen 2.5 14B (2048 input + 256 generation tokens)
8.7End-to-end Latency (s)
4
Qwen VL 2.5
360ASR
4
Qwen VL 7B 2.5
9Throughput (Intel i7-13620H)
3
Qwen (test)
7.06Generation Score
3
Qwen (test)
75.4Accuracy
3
Qwen (test)
9.72Score
3
Qwen3-14B
106Rollout Time
3
Qwen 4B Instruct 3 2507
0BER
3
Qwen 30B on M2 Ultra
0.572Wall Time (s)
3