
Qwen4-Max API
qwen4-maxQwen4 Max (model id qwen4-max) is the newest flagship in Alibaba's Qwen family. Alibaba's Model Studio documentation calls it the next-generation flagship multimodal model of the family and lists it for complex reasoning and coding. On APIMODELS it costs $1.50 per 1M input tokens and $4.50 per 1M output tokens, with cached input at $0.075. That is the same as Alibaba's international list price, so the reason to call Qwen4 Max here is one API key and one USD balance shared with 160+ other models, not a discount. Alibaba lists a single price tier for the whole context window, with no long-context surcharge, and the same rate with thinking on or off. Specifications from Alibaba's documentation: a 1,000,000-token context window, up to 131,072 output tokens per response, a 262,144-token limit shared by thinking and the answer, and text, image and video input with text output. What we verified on APIMODELS on 16 November 2026: the OpenAI-compatible POST /v1/chat/completions endpoint with model qwen4-max, streaming with reasoning_content deltas, function calling, response_format json_schema, and image input by URL. Video input is listed in Alibaba's documentation; we have not tested it through APIMODELS. Thinking is on by default: send enable_thinking: false to turn it off, or thinking_budget to cap it (Alibaba's default budget is 131,072 tokens). Reasoning tokens are counted inside completion_tokens and billed at the output rate, so a short visible answer can still carry a larger output bill; turn thinking off for classification, extraction and other simple high-volume calls. What Qwen4 Max is used for: multi-step coding and code review with a large repository loaded into the 1M window; analysis of long contracts, reports and logs in a single request; agents that call tools and return JSON matching a schema; and questions about screenshots, charts and photos sent as image URLs. Compared with the previous flagship Qwen3.8 Max, Alibaba lists Qwen4 Max at $1.50 / $4.50 where Qwen3.8 Max is $2.00 / $6.00, while the context window (1,000,000 tokens) and the maximum output (131,072 tokens) are the same on both. Alibaba's documentation pages for qwen4-max publish no benchmark scores as of 16 November 2026, so run your own prompts before moving a production workload. Web search is available for Qwen4 Max on the POST /v1/responses endpoint through the web_search tool: each completed search costs $0.012 on top of token charges, and the retrieved content is billed as input tokens at the normal rate. Batch inference is not supported for this model. Billing is pay-per-use with no subscription, and failed calls are not charged.
API Documentation
View complete API reference with all parameters and examples.
Full API Documentation
View complete API reference with streaming, thinking, and more.
Pricing
Billing: Cost = (input_tokens * input_price + output_tokens * output_price) / 1,000,000
Alibaba List Price, One Tier
$1.50 / $4.50 per 1M tokens, cached input $0.075; one rate across the whole 1M window, with thinking on or off
1M Context, 131,072-Token Output
Per Alibaba: 1,000,000-token context window and up to 131,072 output tokens; thinking and answer share 262,144 tokens
Thinking You Can Switch Off
On by default; enable_thinking: false turns it off and thinking_budget caps it. Reasoning tokens bill as output
Tool Calls and JSON Schema
Function calling and response_format json_schema, verified on /v1/chat/completions on 16 Nov 2026; streaming returns reasoning_content
Text and Image Input
Image input by URL verified on APIMODELS; Alibaba also lists video input, which we have not tested here
One Key for 160+ Models
One API key and one USD balance; pay per use with no subscription, and failed calls are not charged
Web Search at $0.012 per Search
On /v1/responses with the web_search tool; charged per completed search on top of tokens, and the response lists the queries and source URLs
Web search with Qwen4 Max on /v1/responses
Qwen4 Max can search the web when you call it on the Responses endpoint with the web_search tool. The request uses input instead of messages, streaming is supported, and the output lists each web_search_call with its query and source URLs before the answer.
A completed search costs $0.012 on top of normal token charges, and the amount is included in usage.cost in the response. The content it retrieves is added to the prompt and billed as input tokens at $1.50 per 1M tokens; in our tests on 16 November 2026 a search typically added a few thousand input tokens.
curl https://api.apimodels.app/v1/responses \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "qwen4-max",
"input": "What changed in the latest Node.js LTS release? Cite your sources.",
"tools": [{"type": "web_search"}]
}'On /v1/responses, Qwen4 Max accepts two tool types: function for your own tools, and web_search. Other built-in tools return a 400 error. Web search is not available on /v1/chat/completions, so send search requests to /v1/responses.
Qwen4 Max vs Qwen3.8 Max: price and limits
Qwen3.8 Max is the previous Qwen flagship and the model most people compare Qwen4 Max with. The rows below are the ones that can be checked against Alibaba's own documentation: list price, context window, output limit, input types and batch support.
| Qwen4 Max | Qwen3.8 Max | |
|---|---|---|
| Alibaba list price, input per 1M tokens | $1.50 | $2.00 |
| Alibaba list price, output per 1M tokens | $4.50 | $6.00 |
| Price on APIMODELS (input / output / cached input, per 1M tokens) | $1.50 / $4.50 / $0.075 | $2.00 / $6.00 / $0.25 |
| Context window | 1,000,000 tokens | 1,000,000 tokens |
| Maximum output per response | 131,072 tokens | 131,072 tokens |
| Thinking limit | 262,144 tokens, shared by thinking and answer | 262,144 tokens (maximum chain-of-thought length) |
| Input types | Text, image, video | Text, image, video |
| Batch inference at Alibaba | Not supported | China (Beijing) region only |
Alibaba figures are from Alibaba Cloud Model Studio documentation (international site), checked 2026-11-16. They are vendor-published, not our own test. Through APIMODELS we have verified text and image-URL input on Qwen4 Max; video input has not been tested here.
There is no quality row because Alibaba's qwen4-max documentation pages publish no benchmark scores as of 2026-11-16, and we do not quote figures from second-hand articles. Context window and maximum output are the same on both models, and batch inference exists only for Qwen3.8 Max. To compare answers, change the model field and run your own prompts on both.
Sources for Qwen4 Max
Model facts and Alibaba prices on this page come from Alibaba Cloud Model Studio documentation, checked 2026-11-16. The APIMODELS prices and the API behaviour we describe as verified are our own.
- Alibaba Cloud Model Studio: qwen4-max model page ↗
Context window, output and thinking limits, input types and supported features. Checked 2026-11-16.
- Alibaba Cloud Model Studio: model pricing ↗
International list prices for qwen4-max ($1.50 / $4.50, cache hit $0.075, one tier) and qwen3.8-max ($2.00 / $6.00). Checked 2026-11-16.
- Alibaba Cloud Model Studio: model list ↗
Where Alibaba lists qwen4-max for complex reasoning and coding, and under visual understanding. Checked 2026-11-16.
- Alibaba Cloud Model Studio: qwen3.8-max model page ↗
The Qwen3.8 Max column of the comparison: context window, output limit, chain-of-thought length and batch availability. Checked 2026-11-16.
About Qwen4 Max
Qwen4 Max is a Large Language Model API provided by Alibaba. Qwen4 Max (model id qwen4-max) is the newest flagship in Alibaba's Qwen family. Alibaba's Model Studio documentation calls it the next-generation flagship multimodal model of the family and lists it for complex reasoning and coding. On APIMODELS it costs $1.50 per 1M input tokens and $4.50 per 1M output tokens, with cached input at $0.075. That is the same as Alibaba's international list price, so the reason to call Qwen4 Max here is one API key and one USD balance shared with 160+ other models, not a discount. Alibaba lists a single price tier for the whole context window, with no long-context surcharge, and the same rate with thinking on or off. Specifications from Alibaba's documentation: a 1,000,000-token context window, up to 131,072 output tokens per response, a 262,144-token limit shared by thinking and the answer, and text, image and video input with text output. What we verified on APIMODELS on 16 November 2026: the OpenAI-compatible POST /v1/chat/completions endpoint with model qwen4-max, streaming with reasoning_content deltas, function calling, response_format json_schema, and image input by URL. Video input is listed in Alibaba's documentation; we have not tested it through APIMODELS. Thinking is on by default: send enable_thinking: false to turn it off, or thinking_budget to cap it (Alibaba's default budget is 131,072 tokens). Reasoning tokens are counted inside completion_tokens and billed at the output rate, so a short visible answer can still carry a larger output bill; turn thinking off for classification, extraction and other simple high-volume calls. What Qwen4 Max is used for: multi-step coding and code review with a large repository loaded into the 1M window; analysis of long contracts, reports and logs in a single request; agents that call tools and return JSON matching a schema; and questions about screenshots, charts and photos sent as image URLs. Compared with the previous flagship Qwen3.8 Max, Alibaba lists Qwen4 Max at $1.50 / $4.50 where Qwen3.8 Max is $2.00 / $6.00, while the context window (1,000,000 tokens) and the maximum output (131,072 tokens) are the same on both. Alibaba's documentation pages for qwen4-max publish no benchmark scores as of 16 November 2026, so run your own prompts before moving a production workload. Web search is available for Qwen4 Max on the POST /v1/responses endpoint through the web_search tool: each completed search costs $0.012 on top of token charges, and the retrieved content is billed as input tokens at the normal rate. Batch inference is not supported for this model. Billing is pay-per-use with no subscription, and failed calls are not charged. Through APIMODELS platform, you can access this model via a unified API with transparent pay-as-you-go pricing. Current pricing: Input: $1.50, Output: $4.50 per 1M tokens.
Key Features
- Alibaba List Price, One Tier -- $1.50 / $4.50 per 1M tokens, cached input $0.075; one rate across the whole 1M window, with thinking on or off
- 1M Context, 131,072-Token Output -- Per Alibaba: 1,000,000-token context window and up to 131,072 output tokens; thinking and answer share 262,144 tokens
- Thinking You Can Switch Off -- On by default; enable_thinking: false turns it off and thinking_budget caps it. Reasoning tokens bill as output
- Tool Calls and JSON Schema -- Function calling and response_format json_schema, verified on /v1/chat/completions on 16 Nov 2026; streaming returns reasoning_content
- Text and Image Input -- Image input by URL verified on APIMODELS; Alibaba also lists video input, which we have not tested here
- One Key for 160+ Models -- One API key and one USD balance; pay per use with no subscription, and failed calls are not charged
- Web Search at $0.012 per Search -- On /v1/responses with the web_search tool; charged per completed search on top of tokens, and the response lists the queries and source URLs
Use Cases
Coding Across a Large Repository
Load source files, tests and docs into the 1,000,000-token window and ask for a multi-file change or a review; one response can run up to 131,072 output tokens. Keep thinking on for multi-step work.
Long Contracts, Reports and Logs
Send the whole document in one request instead of chunking it. Alibaba lists one price tier for the full window, so a long prompt is billed at the same $1.50 per 1M input tokens as a short one.
Tool-Calling Agents
Define functions, let Qwen4 Max decide which to call, and get the final result as JSON that matches your schema through response_format json_schema. Streaming returns reasoning_content while it works.
Questions About Screenshots and Charts
Pass an image URL in the same chat request and ask about a UI screenshot, a chart or a photographed page. Image input by URL was verified on APIMODELS on 16 Nov 2026.
High-Volume Classification and Extraction
For simple calls, send enable_thinking: false so the output bill covers only the answer. Reasoning tokens otherwise count as output at $4.50 per 1M tokens.
Answers Grounded in Live Web Sources
For questions that need current information, call /v1/responses with the web_search tool. Each completed search is $0.012 plus tokens, and the output lists the search queries and source URLs before the answer.
Why APIMODELS
- Unified API -- One API key to access all models, no need to register on multiple platforms
- Transparent Pricing -- Pay-as-you-go with per-call pricing published on every model page, ideal for indie developers and startups
- Instant Access -- Start right after signup; pay by card, PayPal, Alipay or USDT
- Full Documentation -- Detailed API docs with code examples in cURL, Python, and Node.js
Frequently Asked Questions
What is the difference between APIMODELS and the official Alibaba API?
APIMODELS offers the same Qwen4 Max model through our aggregation platform. We provide a unified API interface so you do not need separate accounts for each provider - one API key to access all models.
How much does the Qwen4 Max API cost?
On APIMODELS, $1.50 per 1M input tokens, $4.50 per 1M output tokens and $0.075 per 1M cached input tokens. That is the same as Alibaba's international list price, so there is no discount on this model here; what you get is one API key and one USD balance for 160+ models, pay-per-use billing with no subscription, and no charge for failed calls. Alibaba lists one tier for the whole 1,000,000-token window, so long prompts carry no surcharge, and the rate is the same with thinking on or off; reasoning tokens bill as output. Example: 10,000 input tokens plus 2,000 output tokens cost $0.015 + $0.009 = $0.024.
Alibaba Cloud Model Studio pricing (checked 2026-11-16) ↗Is the Qwen4 Max API free?
No. On APIMODELS, qwen4-max is billed per token from the first call: $1.50 per 1M input tokens and $4.50 per 1M output tokens. There is no subscription and no monthly fee, you pay only for what you use, and failed calls are not charged.
What is the Qwen4 Max context window and maximum output?
Per Alibaba's documentation: a 1,000,000-token context window, up to 131,072 output tokens per response, a 262,144-token limit shared by thinking and the answer, and a default thinking_budget of 131,072. Input is text, image and video; output is text. On APIMODELS we have tested text and image-URL input; video input has not been tested here.
Alibaba Cloud Model Studio, qwen4-max model page (checked 2026-11-16) ↗Qwen4 Max vs Qwen3.8 Max: what is the difference?
Three things can be compared with numbers. Price: Alibaba lists Qwen4 Max at $1.50 / $4.50 per 1M tokens and Qwen3.8 Max at $2.00 / $6.00. Context window: 1,000,000 tokens on both. Maximum output: 131,072 tokens on both. On answer quality, Alibaba's qwen4-max documentation pages publish no benchmark scores as of 16 November 2026, and we do not repeat figures from second-hand articles. Both models sit behind the same endpoint here, so the reliable comparison is to change the model field and run your own prompts.
Alibaba Cloud Model Studio, qwen3.8-max model page (checked 2026-11-16) ↗How do I turn off thinking on Qwen4 Max?
Thinking is on by default. Add enable_thinking: false to the request body to turn it off; to keep thinking but limit it, set thinking_budget. Reasoning tokens are counted inside completion_tokens and billed at the output rate of $4.50 per 1M tokens, so switching thinking off for simple work such as classification or extraction cuts output tokens noticeably. With streaming, the reasoning arrives as reasoning_content deltas. This is the behaviour we measured on APIMODELS on 16 November 2026.
How do I call the Qwen4 Max API with the OpenAI SDK?
The endpoint is OpenAI-compatible: POST https://api.apimodels.app/v1/chat/completions with Authorization: Bearer YOUR_API_KEY and body {"model":"qwen4-max","messages":[{"role":"user","content":"..."}]}. With openai-python or openai-node, set base_url to https://api.apimodels.app/v1 and keep the rest of your code. Add stream: true for streaming. We verified streaming, function calling and response_format json_schema on 16 November 2026.
Does Qwen4 Max support image input and function calling?
Yes. Images are passed by URL inside messages, which we verified on APIMODELS, along with function calling and response_format json_schema. Alibaba's documentation also lists parallel tool calls, JSON Object output, implicit and explicit context caching, and video input; we have not tested video input or explicit caching with cache_control through APIMODELS, so check them yourself before relying on them. Cached input is billed at $0.075 per 1M tokens. Batch inference is not supported for this model.
Are there Qwen4 Max benchmarks?
We checked Alibaba's Model Studio pages for qwen4-max on 16 November 2026: the model page, the pricing page and the model list. They publish specifications and prices, and no benchmark scores. We do not pass on numbers from second-hand articles. Alibaba positions the model for complex reasoning and coding; to judge it for your workload, run your own prompts against the previous flagship Qwen3.8 Max, which uses the same endpoint with only the model field changed.
Can Qwen4 Max search the web through the API?
Yes, on the Responses endpoint. Send POST https://api.apimodels.app/v1/responses with model qwen4-max, your question in input (this endpoint takes input instead of messages) and tools: [{"type":"web_search"}]; streaming is supported. The output contains web_search_call items with the query and the source URLs, followed by the answer. Each completed search costs $0.012 on top of normal token charges, and that amount is included in usage.cost in the response. The retrieved content is added to the prompt and billed as input tokens at $1.50 per 1M tokens; in our tests on 16 November 2026 a search typically added a few thousand input tokens. On /v1/responses this model accepts two tool types, function for your own tools and web_search; other built-in tools return a 400 error. Web search is not available on /v1/chat/completions, so use /v1/responses for it.
What can Qwen4 Max do?
Qwen4 Max supports: Reasoning / Thinking, 1M Context, 131K Output, Tool Calling, JSON Schema, Image Input. See the APIMODELS docs for full parameters and call examples.
Can I use the Qwen4 Max API from any country, including mainland China?
Yes. APIMODELS does not restrict access by country: the site and the API answer requests from anywhere, and apimodels.app and api.apimodels.app are reachable directly from mainland China without a proxy. You call Qwen4 Max with one APIMODELS key, so you do not need a separate account with the model's developer.
What payment methods are supported? Can I pay with PayPal or crypto?
Yes to both. You can top up with a card (Visa, Mastercard and others, through Stripe), PayPal, Alipay, or USDT on TRON (TRC20) or BNB Smart Chain (BEP20). Card, PayPal and Alipay top-ups are credited immediately; USDT is credited once the transfer confirms on-chain. Auto top-up is optional, and a $50 top-up adds a $1 bonus ($3 on a $100 top-up). The balance is in USD and billed per use, with no subscription.
Related models
Related reading
How to get access, regional availability, and how this model compares with its alternatives.
- Qwen 4 Max Review: Benchmarks, Price, Same-Prompt Tests and Where It Falls Short
- Qwen 4 Max vs GPT-6 Sol: Benchmarks, Same-Prompt Tests and Cost per Task
- Qwen 4 Max vs Gemini 4 Carbon: Benchmarks, Same-Prompt Tests, Price and Speed
- Qwen 4 Max vs Claude Opus 5.5: Benchmarks, Coding Tests and Cost per Task
- Use Claude, GPT and Qwen in Cline with One API Key: OpenAI Compatible Setup
- Add a Custom API Provider in Cherry Studio: Claude, GPT and Qwen with One Key
- Add a Custom AI Provider in LobeHub: Claude, GPT and Qwen with One Key