Alibaba
Qwen3 VL 32B Instruct
Qwen3-VL-32B-Instruct is a large-scale multimodal vision-language model designed for high-precision understanding and reasoning across text, images, and video. With 32 billion parameters, it combines deep visual perception with advanced text...
Qwen model releasesModel Snapshot
ReleaseOct 21, 2025
Context131,072
Overall
Rank #343 across 566
11.1
Coding
No rank
n/a
Blended Price
Rank #201 across 371
$1.23/M
Strength Profile
Percentile score by analysis domain.
* Cost is inverted: lower input, output, and blended prices rank higher.
Benchmark Percentiles
Higher bars mean stronger relative placement.
Model Details
Runtime
Output Speed80.6 tok/s
First Token1.07s
Max Output32,768
Pricing
Blended$1.23/M
Input$0.700/M
All Benchmarks
| Metric | Domain | Value | Rank |
|---|---|---|---|
| Artificial Analysis Intelligence Index | overall | 11.1 | #343 |
| Artificial Analysis Math Index | math | 68.3 | #100 |
| MMLU-Pro | reasoning | 79.1% | #136 |
| GPQA | reasoning | 67.1% | #281 |
| Humanity's Last Exam |