Back to Explore
Qwen: Qwen3 VL 235B A22B Instruct
Alibaba·Released 2025-09-23
Open Source235B262K ctxApache 2.0Multimodal
About
Qwen3-VL-235B-A22B Instruct is an open-weight multimodal model that unifies strong text generation with visual understanding across images and video. The Instruct model targets general vision-language use (VQA, document parsing, chart/table...
Related Models
Pricing
Input
$0.20
per 1M tokens
Output
$0.88
per 1M tokens
Blended
$0.37
per 1M tokens
Cheaper than 58% of models. Median price is $0.54/1M tokens.
Cost Calculator
Tokens per day1M
100K100M
Daily
$0.37
Monthly
$11.10
vs. Similar Models
GPT-5 mini (minimal)Q:0.0
$0.69+86%
Meta: Llama 4 MaverickQ:0.0
$0.26-29%
gpt-oss-20B (low)Q:0.0
$0.10-74%
Nova 2.0 Pro Preview (Non-reasoning)Q:+0.1
$3.44+829%
Performance
50
tokens/sec
Faster than 16% of models
1.18
seconds
Faster than 45% of models
1.18
seconds
Faster than 63% of models
Market Median
94 tok/s
47% slower
Median TTFT
1.11s
6% slower
Throughput/Dollar
135
tok/s per $/1M
Speed Comparison
Mistral: Mistral Medium 3
50 tok/s-0%
MiniMax: MiniMax M2.7
50 tok/s+0%
Z.ai: GLM 4.6
50 tok/s+1%
Context Window
262K
tokens
Larger than 62% of models
Max Output
16K
tokens
6% of context
Benchmarks
MMLU-Pro
82.3%
GPQA Diamond
71.2%
HLE
6.3%
LiveCodeBench
59.4%
SciCode
35.9%
TerminalBench Hard
6.8%
MATH-500Not evaluated
AIMENot evaluated
AIME 2025
70.7%
IFBench
42.7%
Long Context Recall
31.7%
Tau2
35.1%
Market AverageTop Score
Open Source
apache-2.0235BGGUF / GPTQ / AWQ
Downloads
1.7M
Likes
398
VRAM (FP16)
Multi-GPU
GPU
8x A100 / H100