Text generation and visual understanding

​

Choose a model

Most Qwen models accept text, images, and video as input and answer with text. Use the same model to generate content, understand images, or analyze video — you do not need to switch to a VL model just because the input contains images or video.

• Default: qwen3.7-plus — balanced performance and cost.
• Highest quality: qwen4-max-preview (Token Plan only) — for tasks that demand stronger reasoning and understanding.
• Lowest cost: qwen3.8-flash — confirm that quality meets your needs before production use.

See the model tables below for input modalities and feature support.

​

Using OpenClaw or Claude Code?

qwen3.7-plus — balanced performance and cost, full tool support, 1M context for large codebases. For advanced reasoning, choose the flagship qwen4-max-preview (Token Plan only). If you prefer third-party models, choose kimi-k3 for long-horizon coding or glm-5.3 for open-source SOTA coding and agent tasks.

​

Migrate from closed-source models

Map your current GPT, Claude, or Gemini model to an equivalent QwenCloud model. All models recommended above also support image and video understanding, so you can use them for image analysis, video understanding, and OCR.

​

For other applications

Chatbots, content generation, summarization, document processing, meeting notes, and data analytics — start with qwen3.7-plus, which balances performance and cost with a 1M-token context window, Function Calling support, and built-in tools. To cut costs, switch to qwen3.8-flash, which delivers near-flagship performance at a lower price with the same context length. For the advanced reasoning, such as complex data analysis and multi-step logic, use the flagship qwen4-max-preview (Token Plan only). When a document exceeds 1 million tokens, import it into a knowledge base, or process it in segments.

​

Context window

1M tokens is roughly 1.5 million Chinese characters.

• Long documents or large codebases → qwen4-max-preview (Token Plan only), qwen3.8-max / qwen3.7-max / qwen3.7-plus / qwen3.8-flash (1M)
• Third-party models such as kimi-k3 / glm-5.3 / deepseek-v4.1-flash-1017 also offer a 1M-token context window
• Standard tasks → 128k–256k is plenty

​

Thinking mode

Step-by-step reasoning for multi-step math, debugging, architecture planning, or legal cross-referencing. Toggle with enable_thinking, or use reasoning.effort in the Responses API to control thinking depth. All Qwen3+ models support it — most are hybrid, so you can switch per request.

​

Function calling + built-in tools

Let the model take actions: check weather, query a database, book a meeting.

• Function calling (you define tools, model calls them): all general-purpose models
• Built-in tools (web search, code execution — no setup): qwen4-max-preview (Token Plan only), qwen3.8-max, qwen3.8-max-0902, qwen3.8-flash, qwen3.7-max, qwen3.7-plus, qwen3.7-flash, qwen3.6-flash, qwen3.5-plus, qwen3.5-flash

​

Image and video understanding

Start with qwen4-max-preview (Token Plan only) — a Qwen flagship model, 1M context, 2-hour video support, and the full feature set including function calling and built-in tools. Once your use case works well, try qwen3.7-plus or qwen3.8-flash to reduce cost — near-flagship quality with the same context and features.

​

Image resolution

Most models support up to 16M pixels per image. Higher resolution costs more tokens: each image uses h × w / (32 × 32) + 2 tokens.

​

Video support

• Up to 2 hours / 2GB → qwen4-max-preview (Token Plan only), qwen3.8-max, qwen3.8-max-0902, qwen3.8-flash, qwen3.8-omni-flash, qwen3.7-plus, qwen3.7-flash, qwen3.6-plus, qwen3.6-flash, qwen3.5-plus, qwen3.5-flash
• Up to 1 hour / 2GB → qwen3.5-omni-plus, qwen3.5-omni-flash (also accepts audio — see Speech models)

To understand sound in a video or provide audio input, use qwen3.8-omni-flash, which produces text output. See Qwen3.8-Omni-Flash for invocation examples.

​

Function calling + built-in tools

Let the model take actions based on what it sees in images or video.

• Function calling: Qwen3.8, Qwen3.7, Qwen3.6, Qwen3.5, Qwen3.5-Omni (including realtime models), and Qwen3-VL models
• Built-in tools (web search, code execution — no setup): qwen4-max-preview (Token Plan only), qwen3.8-max, qwen3.8-max-0902, qwen3.8-flash, qwen3.7-plus, qwen3.7-flash, qwen3.6-plus, qwen3.6-flash, qwen3.5-plus, qwen3.5-flash only. qwen3.8-omni-flash supports the web_search built-in tool.

​

Structured output

Get valid JSON from visual input — e.g., extract product info from a photo. Available on Qwen3.8, Qwen3.7, Qwen3.6, and Qwen3.5 in non-thinking mode. Qwen3.8-Omni-Flash supports JSON Object mode in both thinking and non-thinking modes.

​

Text extraction

For text extraction, qwen3.5-ocr supports document parsing, text localization, key information extraction, multi-turn conversations, and PDF parsing. It is available in China North 2 (Beijing) only. See Text extraction.

​

Recommended models

​

Text generation

​

Image and video parameters

​

All models

​

Text generation

The model pages in the console, including model details and model comparison, show the models, prices, and rate limits of the region selected in the upper-right corner. A model offered in one region may not be listed in another. If you switch regions while a model detail page or the comparison list is open, a model that is not available in the new region is flagged, and the page shows the other regions where the model is available. The comparison page stays open even if only one model remains.

​

Visual understanding

The model pages in the console, including model details and model comparison, show the models, prices, and rate limits of the region selected in the upper-right corner. A model offered in one region may not be listed in another. If you switch regions while a model detail page or the comparison list is open, a model that is not available in the new region is flagged, and the page shows the other regions where the model is available. The comparison page stays open even if only one model remains.

---

​

Learn more