Back to Explore
Qwen3 VL 8B Instruct
Alibaba·Released 2025-10-11
Open Source8B256K ctxApache 2.0Multimodal
About
Qwen3-VL-8B-Instruct is a multimodal vision-language model from the Qwen3-VL series, built for high-fidelity understanding and reasoning across text, images, and video. It features improved multimodal fusion with Interleaved-MRoPE for long-horizon...
Related Models
Pricing
Input
$0.08
per 1M tokens
Output
$0.50
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
$5.55
vs. Similar Models
Qwen3 4B (Reasoning)Q:0.0
$0.40+115%
Llama 3.1 Instruct 405BQ:+0.1
$3.69+1894%
Claude 3.5 Sonnet (June '24)Q:-0.1
$6.00+3143%
Llama 3.3 Instruct 70BQ:+0.2
$0.61+231%
Performance
143
tokens/sec
Faster than 70% of models
0.94
seconds
Faster than 58% of models
0.94
seconds
Faster than 71% of models
Market Median
94 tok/s
51% faster
Median TTFT
1.11s
15% faster
Throughput/Dollar
771
tok/s per $/1M
Speed Comparison
Sarvam M (Reasoning)
143 tok/s-0%
GPT-5 nano (medium)
142 tok/s-0%
Google: Gemini 2.5 Pro
142 tok/s-0%
Context Window
256K
tokens
Larger than 58% of models
Max Output
33K
tokens
13% of context
Benchmarks
MMLU-Pro
68.6%
GPQA Diamond
42.7%
HLE
2.9%
LiveCodeBench
33.2%
SciCode
17.4%
TerminalBench Hard
2.3%
MATH-500Not evaluated
AIMENot evaluated
AIME 2025
27.3%
IFBench
32.3%
Long Context Recall
15.3%
Tau2
29.2%
Market AverageTop Score
Open Source
apache-2.08BGGUF / GPTQ / AWQ
Downloads
5.2M
Likes
975
VRAM (FP16)
8-16 GB
GPU
RTX 4070 / M2 Pro