Choose a Model

Model Comparison

Model	Context	Input	Output	Strengths
qwen3-max	262K	$1.20/1M	$6.00/1M	Best reasoning, complex coding, research
qwen3.5-plus	1M	$0.40/1M	$2.40/1M	Long documents, RAG, balanced quality
qwen3.5-flash	1M	$0.10/1M	$0.40/1M	Fast responses, high-volume pipelines
qwen-turbo	1M	$0.05/1M	$0.20/1M	Cheapest option, simple tasks
qwen-vl-max	—	$0.80/1M	$3.00/1M	Image understanding, OCR, charts
qwen3-omni	—	—	—	Audio, video, and text in one model

Decision Framework

Start with qwen3.5-flash for most new projects. It's fast, cheap, and handles the majority of chat and text tasks well. Move up or down from there based on your results.

• Need the best output quality? → qwen3-max
• Processing long documents or large codebases? → qwen3.5-plus (1M context)
• Running at scale and watching costs? → qwen-turbo
• Working with images or screenshots? → qwen-vl-max
• Need audio or video understanding? → qwen3-omni

Thinking Mode

qwen3-max and qwen3.5-plus support an extended thinking mode that improves accuracy on hard reasoning tasks — math, logic, multi-step planning. Enable it by passing enable_thinking: true in the extra body:

``python

response = client.chat.completions.create(

model="qwen3-max",

messages=[{"role": "user", "content": "Solve: if 3x + 7 = 22, what is x?"}],

extra_body={"enable_thinking": True},

)

`

Thinking mode uses more tokens and is slower. Use it when accuracy matters more than speed or cost.

Model Versioning

Model names without a date suffix (e.g., qwen3-max) always point to the latest stable version. If you need reproducible outputs, pin to a dated snapshot:

`

qwen3-max-2025-09-19

qwen3.5-flash-2025-11-01

`

Dated snapshots are supported for at least 12 months after release. Check the changelog for available snapshots.

Context Length Notes

qwen3-max` has a 262K context window — enough for most codebases and long documents. The other text models support up to 1M tokens, which fits roughly 750,000 words or a full novel.

Longer contexts cost more. If your use case doesn't need the full window, truncate your input to keep costs down.