Qwen3 VL 32B Instruct
by Qwentext+image->text1 endpoint
Neura Intelligence Index
39.5/ 100
Rank
#197—
Confidence
mediumBenchmarks
6(15% coverage)
Overview
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...
Capabilities
Text generation
Image understanding
Tool use / Function calling
Structured output
Long context
Top-P sampling
Modalities
Input
TextImage
Output
Text
Technical Specifications
Context Window
262.1K tokens
Max Output
32.8K tokens
Tokenizer
Qwen
Uptime (24h)
10000.0%
Parameters
33000000000B
License
apache_2_0Open Weights
Supported Parameters
logprobsmax_tokenspresence_penaltyresponse_formatseedstructured_outputstemperaturetool_choicetoolstop_logprobstop_p
Benchmark Performance
math
| Benchmark | Score | Source |
|---|---|---|
| AIME 2025 | 66.2 | verified |
reasoning
| Benchmark | Score | Source |
|---|---|---|
| GPQA Diamond | 68.9 | verified |
multimodal
| Benchmark | Score | Source |
|---|
Performance Over Time
Pricing
Per 1M tokens
Input$0.10
Output$0.42