Back to Explore
Qwen: Qwen3 VL 32B Instruct
Alibaba·Released 2025-10-23
Open Source32B262K ctxApache 2.0Multimodal
About
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...
Related Models
Pricing
Input
$0.10
per 1M tokens
Output
$0.42
per 1M tokens
Blended
$0.18
per 1M tokens
Cheaper than 71% of models. Median price is $0.54/1M tokens.
Cost Calculator
Tokens per day1M
100K100M
Daily
$0.18
Monthly
$5.46
vs. Similar Models
Ministral 3 14BQ:0.0
$0.20+10%
OpenAI: GPT-4oQ:+0.1
$4.38+2304%
DeepSeek R1 Distill Qwen 32BQ:-0.1
$0.29+59%
Z.ai: GLM 4.6VQ:-0.1
$0.45+147%
Performance
72
tokens/sec
Faster than 37% of models
1.10
seconds
Faster than 52% of models
1.10
seconds
Faster than 66% of models
Market Median
94 tok/s
24% slower
Median TTFT
1.11s
1% faster
Throughput/Dollar
395
tok/s per $/1M
Speed Comparison
Nova Premier
72 tok/s-0%
OpenAI: GPT-4o-mini
73 tok/s+1%
NVIDIA Nemotron Nano 9B V2 (Reasoning)
71 tok/s-1%
Context Window
262K
tokens
Larger than 62% of models
Max Output
33K
tokens
13% of context
Benchmarks
MMLU-Pro
79.1%
GPQA Diamond
67.1%
HLE
6.3%
LiveCodeBench
51.4%
SciCode
30.1%
TerminalBench Hard
8.3%
MATH-500Not evaluated
AIMENot evaluated
AIME 2025
68.3%
IFBench
39.2%
Long Context Recall
31.3%
Tau2
29.2%
Market AverageTop Score
Open Source
apache-2.032BGGUF / GPTQ / AWQ
Downloads
5.4M
Likes
207
VRAM (FP16)
24-48 GB
GPU
A6000 / M3 Ultra