Alibaba Cloud / Qwen Team

Qwen2.5 7B Instruct

Zero-eval
#3AlignBench

by Alibaba Cloud / Qwen Team

+
+
+
+
About

Qwen 2.5 7B was created as an efficient instruction-tuned model, designed to provide capable performance with just 7 billion parameters. Built for applications requiring reliable language understanding with minimal computational overhead, it serves as an accessible entry point to the Qwen 2.5 family.

+
+
+
+
Pricing Range
Input (per 1M)$0.30 -$0.30
Output (per 1M)$0.30 -$0.30
Providers1
+
+
+
+
Timeline
AnnouncedSep 19, 2024
ReleasedSep 19, 2024
+
+
+
+
Specifications
Training Tokens18.0T
+
+
+
+
License & Family
License
Apache 2.0
Performance Overview
Performance metrics and category breakdown

Overall Performance

14 benchmarks
Average Score
65.6%
Best Score
91.6%
High Performers (80%+)
3

Performance Metrics

Max Context Window
139.3K
Avg Throughput
138.0 tok/s
Avg Latency
1ms
+
+
+
+
All Benchmark Results for Qwen2.5 7B Instruct
Complete list of benchmark scores with detailed information
GSM8k
text
0.92
91.6%
Self-reported
MT-Bench
text
0.88
87.5%
Self-reported
HumanEval
text
0.85
84.8%
Self-reported
MBPP
text
0.79
79.2%
Self-reported
MATH
text
0.76
75.5%
Self-reported
MMLU-Redux
text
0.75
75.4%
Self-reported
AlignBench
text
0.73
73.3%
Self-reported
IFEval
text
0.71
71.2%
Self-reported
MultiPL-E
text
0.70
70.4%
Self-reported
MMLU-Pro
text
0.56
56.3%
Self-reported
Showing 1 to 10 of 14 benchmarks
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+