All AI models · LLM benchmarks · Methodology

Alibaba

Qwen2.5 VL 72B

Qwen vision-language model for visual reasoning, documents, and agent tasks

Model facts

Context window
128K tokens
Maximum output
8K tokens
Cheapest paid input
$0.25 per 1M tokens
Cheapest paid output
$0.747 per 1M tokens
Open weights
yes
Providers
8

Capabilities and modalities

reasoning, tool calling, structured output, attachments, temperature control, open weights, text, image, audio, video.

Local hardware estimate

At 4K context: Q4_K_M 45.9 GiB, Q8_0 80.1 GiB, F16 144.2 GiB working memory.

Parameters
73.4 billion
Architecture
qwen2_5_vl
Model license
other

Planning estimate adapted from the MIT-licensed whichllm estimator using metadata from Hugging Face; not a vendor minimum requirement.

Observed price history

  • 2026-08-24: $0.25 input / $0.747 output per 1M tokens

API providers

  • Cortecs — model id qwen2.5-vl-72b-instruct; $0.25 input / $0.747 output per 1M tokens; provider documentation
  • NanoGPT — model id qwen25-vl-72b-instruct; $0.7 input / $0.7 output per 1M tokens; provider documentation
  • NovitaAI — model id qwen/qwen2.5-vl-72b-instruct; $0.8 input / $0.8 output per 1M tokens; provider documentation
  • OpenRouter — model id qwen/qwen2.5-vl-72b-instruct; $0.8 input / $1 output per 1M tokens; provider documentation
  • OVHcloud AI Endpoints — model id qwen2.5-vl-72b-instruct; $1.01 input / $1.01 output per 1M tokens; provider documentation
  • Alibaba (China) — model id qwen2-5-vl-72b-instruct; $2.29 input / $6.88 output per 1M tokens; provider documentation
  • Alibaba — model id qwen2-5-vl-72b-instruct; $2.8 input / $8.4 output per 1M tokens; provider documentation
  • Qiniu — model id qwen2.5-vl-72b-instruct; not published input / not published output per 1M tokens; provider documentation