Text analysis and question answering
- Tool calling
支持工具调用 · 支持
- Reasoning
支持推理 · 支持
- Vision
支持视觉 · 支持
- Streaming
支持流式输出 · 支持
- JSON mode
支持JSON 模式 · 支持
USD / 1M tokens
- Input
- $1.50
- Cache read
- $0.075
- Cache write
- $1.875
- Output
- $4.50
- Context
- 983,616
- Max output
- 256K
- Precision
- 官方精度
- Price checked
Model availability and prices
Overview
Qwen4-Max is a Qwen model served through Alibaba Cloud Model Studio. This route accepts text and images and returns text, with switchable reasoning, tool calls and text-only structured output. The official context is 1,000,000 tokens and the combined thinking/answer output maximum is 262,144 tokens (256K), while non-thinking output is limited to 131,072 tokens (128K); the gateway uses a conservative configured limit of 983,616 tokens. qwen4-max is a rolling alias, with no immutable snapshot currently listed.
- Version
- qwen4-max (rolling alias)
- Input / output
- text, image / text
- Endpoint
- /v1/chat/completions, /v1/responses, /v1/messages
API protocols and integration
- Base URL · OpenAI SDK
https://api.iteracompute.com/v1- 鉴权
Authorization: Bearer ITERACOMPUTE_API_KEY
按接口选择请求字段、读取结果和流式结束标记;下表是协议格式,实际能力与验证范围以本模型说明为准。
| 客户接口 | 本模型状态 | 输入 / 输出预算 | 返回结果 | SSE 结束标记 |
|---|---|---|---|---|
OpenAI Chat CompletionsPOST /v1/chat/completions | 已列入模型资料 | messagesmax_tokens | choices[0].message.content | [DONE] |
OpenAI ResponsesPOST /v1/responses | 已列入模型资料 | inputmax_output_tokens | output[] (SDK: output_text) | response.completed |
Anthropic MessagesPOST /v1/messages | 已列入模型资料 | messages; system (top-level)max_tokens (required) | content[] | message_stop |
Messages SDK 的 Base URL 使用 https://api.iteracompute.com;max_tokens 为必填。其他协议字段请勿直接套用。
LLMs.txt · 复制给 AI 编程助手
包含此模型的 ID、接口格式和参数约束。复制前请核对本页验证范围与当前价格。
# IteraCompute · qwen/qwen4-max
Model ID: qwen/qwen4-max
Base URL (OpenAI SDK): https://api.iteracompute.com/v1
Authentication: Authorization: Bearer ITERACOMPUTE_API_KEY
Content-Type: application/json
仅使用下列客户接口;不同协议的请求和响应结构不可混用。接口声明不等于每项能力均已实测。
## Chat Completions
POST https://api.iteracompute.com/v1/chat/completions
Input: messages
Output budget: max_tokens
Read result: choices[0].message.content
SSE completion: [DONE]
Docs: https://iteracompute.com/docs.html#api-reference
## Responses
POST https://api.iteracompute.com/v1/responses
Input: input
Output budget: max_output_tokens
Read result: output[] (SDK: output_text)
SSE completion: response.completed
Docs: https://iteracompute.com/docs.html#responses-api
## Messages
POST https://api.iteracompute.com/v1/messages
Input: messages; system (top-level)
Output budget: max_tokens (required)
Read result: content[]
SSE completion: message_stop
Docs: https://iteracompute.com/docs.html#anthropic-messages
## Parameters
配置范围和参数可接收性不等于模型语义能力验收;未公布的默认值不要自行补全。
Parameter revision: 7e613f33e3eb8fe978e5e155b9783070df559968b8342b443854839aa948729f
Context tokens: 983616
Maximum output tokens: 262144
Capability tools: configured enabled (see verification scope)
Capability reasoning: configured enabled (see verification scope)
Capability vision: configured enabled (see verification scope)
Capability streaming: configured enabled (see verification scope)
Capability json_mode: configured enabled (see verification scope)
- max_tokens: integer; accepted; min=1; max=262144; default=未公布. 输出 Token 预算;适用时包含推理 Token。用于 Chat Completions / Messages。
- max_completion_tokens: integer; accepted; min=1; max=262144; default=未公布. Chat 输出预算的另一字段,仅在此模型接收时使用。
- max_output_tokens: integer; accepted; min=1; max=262144; default=未公布. Responses 输出预算;Chat 是否接收此字段须单独核对。
- stream: boolean; accepted; min=unknown; max=unknown; default=未公布. 启用 SSE 流式输出;结束标记因协议而异。
- tools: array; accepted; min=unknown; max=unknown; default=未公布. 工具定义;由应用执行工具调用并回传结果。
- tool_choice: any; accepted; min=unknown; max=unknown; default=未公布. 控制是否调用以及调用哪个工具;值的结构因协议而异。
- response_format: object; accepted; min=unknown; max=unknown; default=未公布. 文本:Chat 输出格式。图片:只接收 b64_json,与文本 JSON 模式不同。
- reasoning: any; accepted; min=unknown; max=unknown; default=未公布. 推理控制;对象或布尔值的具体用法取决于模型和协议。
- reasoning_effort: enum; accepted; values=["none","low","high","xhigh"]; default=high. Chat 推理强度;使用此模型列出的枚举值,不混用其他协议的结构。
## Chat Completions example · JSON body
```json
{
"model": "qwen/qwen4-max",
"messages": [
{
"role": "user",
"content": "Reply OK"
}
],
"max_tokens": 256,
"reasoning_effort": "low"
}
```
## Responses example · JSON body
```json
{
"model": "qwen/qwen4-max",
"input": "Reply OK",
"max_output_tokens": 256,
"reasoning": {
"effort": "low"
}
}
```
## Messages example · JSON body
```json
{
"model": "qwen/qwen4-max",
"max_tokens": 256,
"messages": [
{
"role": "user",
"content": "Reply OK"
}
],
"output_config": {
"effort": "low"
}
}
```
## References
Model catalog: https://iteracompute.com/models/
Current prices and billing: https://iteracompute.com/pricing/
Documentation index: https://iteracompute.com/llms.txt
Unknown prices are not zero. Account and subscription limits also apply.
Never include a real API key in an LLM prompt or client-side code.
API parameters
- Customer endpoints include OpenAI Chat Completions, OpenAI Responses and Anthropic Messages, using native fields for each protocol.
- Formal model output budgets are 262,144 tokens for thinking and answer combined, or 131,072 with thinking disabled. These are documented limits; accepting a large budget is not full-length generation verification.
- Chat response_format and Responses text.format map json_object to an open object JSON Schema. Explicit customer JSON Schemas are preserved.
- Reasoning can be disabled; verified efforts are low, high and xhigh. Messages uses thinking and output_config.effort, without an exact thinking.budget_tokens control.
- Chat streams end with [DONE], Responses with response.completed, and Messages with message_stop.
- Retail follows the standard Singapore international rate: input $1.50, cache read $0.075, cache write $1.875 at 5 minutes / $3.00 at 1 hour, and output $4.50, all per million tokens.
以下为当前线路接受的参数与取值范围,账户额度和上下文限制同时适用。参数可接收不代表相关模型能力已经完成验证。
- 工具调用
- 支持
- 推理
- 支持
- 视觉
- 支持
- 流式输出
- 支持
- JSON 模式
- 支持
参数与取值
点击参数名即可复制;表格可横向滚动查看全部列。用途栏标明协议差异,不能将同名字段视为跨接口通用。
| 参数 | 类型 | 接收状态 | 取值范围 | 用途 / 协议 |
|---|---|---|---|---|
| 整数 | 可接收 | [1, 262144] | 输出 Token 预算;适用时包含推理 Token。用于 Chat Completions / Messages。 | |
| 整数 | 可接收 | [1, 262144] | Chat 输出预算的另一字段,仅在此模型接收时使用。 | |
| 整数 | 可接收 | [1, 262144] | Responses 输出预算;Chat 是否接收此字段须单独核对。 | |
| 布尔值 | 可接收 | — | 启用 SSE 流式输出;结束标记因协议而异。 | |
| 数组 | 可接收 | — | 工具定义;由应用执行工具调用并回传结果。 | |
| 按参数约定 | 可接收 | — | 控制是否调用以及调用哪个工具;值的结构因协议而异。 | |
| 对象 | 可接收 | — | 文本:Chat 输出格式。图片:只接收 b64_json,与文本 JSON 模式不同。 | |
| 按参数约定 | 可接收 | — | 推理控制;对象或布尔值的具体用法取决于模型和协议。 | |
| 枚举值 | 可接收 | ["none","low","high","xhigh"] | Chat 推理强度;使用此模型列出的枚举值,不混用其他协议的结构。 |
当前配置调用示例
POST /v1/chat/completions
{
"model": "qwen/qwen4-max",
"messages": [
{
"role": "user",
"content": "Reply OK"
}
],
"max_tokens": 256,
"reasoning_effort": "low"
}参数版本:7e613f33e3eb
API
import os
from openai import OpenAI
client = OpenAI(
base_url="https://api.iteracompute.com/v1",
api_key=os.environ["ITERACOMPUTE_API_KEY"],
)
response = client.chat.completions.create(
model="qwen/qwen4-max",
messages=[
{
"role": "user",
"content": "Reply OK"
}
],
max_tokens=256,
reasoning_effort="low",
)
print(response.choices[0].message.content)import OpenAI from "openai";
const client = new OpenAI({
baseURL: "https://api.iteracompute.com/v1",
apiKey: process.env.ITERACOMPUTE_API_KEY,
});
const response = await client.chat.completions.create({
"model": "qwen/qwen4-max",
"messages": [
{
"role": "user",
"content": "Reply OK"
}
],
"max_tokens": 256,
"reasoning_effort": "low"
});
console.log(response.choices[0].message.content);curl https://api.iteracompute.com/v1/chat/completions \
-H "Authorization: Bearer ITERACOMPUTE_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "qwen/qwen4-max",
"messages": [
{
"role": "user",
"content": "Reply OK"
}
],
"max_tokens": 256,
"reasoning_effort": "low"
}'import os
from openai import OpenAI
client = OpenAI(
base_url="https://api.iteracompute.com/v1",
api_key=os.environ["ITERACOMPUTE_API_KEY"],
)
response = client.responses.create(
model="qwen/qwen4-max",
input="Reply OK",
max_output_tokens=256,
reasoning={
"effort": "low"
},
)
print(response.output_text)import OpenAI from "openai";
const client = new OpenAI({
baseURL: "https://api.iteracompute.com/v1",
apiKey: process.env.ITERACOMPUTE_API_KEY,
});
const response = await client.responses.create({
"model": "qwen/qwen4-max",
"input": "Reply OK",
"max_output_tokens": 256,
"reasoning": {
"effort": "low"
}
});
console.log(response.output_text);curl https://api.iteracompute.com/v1/responses \
-H "Authorization: Bearer ITERACOMPUTE_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "qwen/qwen4-max",
"input": "Reply OK",
"max_output_tokens": 256,
"reasoning": {
"effort": "low"
}
}'import os
from anthropic import Anthropic
client = Anthropic(
base_url="https://api.iteracompute.com",
api_key=os.environ["ITERACOMPUTE_API_KEY"],
)
response = client.messages.create(
model="qwen/qwen4-max",
max_tokens=256,
messages=[
{
"role": "user",
"content": "Reply OK"
}
],
output_config={
"effort": "low"
},
)
print("".join(block.text for block in response.content if block.type == "text"))import Anthropic from "@anthropic-ai/sdk";
const client = new Anthropic({
baseURL: "https://api.iteracompute.com",
apiKey: process.env.ITERACOMPUTE_API_KEY,
});
const response = await client.messages.create({
"model": "qwen/qwen4-max",
"max_tokens": 256,
"messages": [
{
"role": "user",
"content": "Reply OK"
}
],
"output_config": {
"effort": "low"
}
});
console.log(response.content.filter(block => block.type === "text").map(block => block.text).join(""));curl https://api.iteracompute.com/v1/messages \
-H "x-api-key: ITERACOMPUTE_API_KEY" \
-H "anthropic-version: 2023-06-01" \
-H "Content-Type: application/json" \
-d '{
"model": "qwen/qwen4-max",
"max_tokens": 256,
"messages": [
{
"role": "user",
"content": "Reply OK"
}
],
"output_config": {
"effort": "low"
}
}'Pricing
USD per million tokens. — means unknown. Account and subscription limits still apply.
- Version / Model ID
qwen/qwen4-maxqwen4-max (rolling alias)- Context
- 983,616
- Precision
- 官方精度
- Price checked
- Verification scope and sources
View verification scope and sources
2026-11-17 · Verified three API formats, ordinary/streaming usage, JSON, function calls, observable reasoning and two-color image recognition. Budget admission tested separately for thinking (262144) and non-thinking (131072). Synthetic input admission reached 655,405 tokens; full-window quality, full-length output and concurrency were not measured. Cache creation/control and stored conversations were not verified.
- help.aliyun.comhelp.aliyun.com/zh/model-studio/qwen4-max
- help.aliyun.comhelp.aliyun.com/en/model-studio/qwen4-max
- modelstudio.console.alibabacloud.commodelstudio.console.alibabacloud.com/ap-southeast-1/model/market/detail/qwen4-max
- docs.modelstudio.console.alibabacloud.comdocs.modelstudio.console.alibabacloud.com/zh/model-studio/qwen-api-via-openai-chat-completions
- docs.modelstudio.console.alibabacloud.comdocs.modelstudio.console.alibabacloud.com/zh/model-studio/qwen-api-via-openai-responses
- docs.modelstudio.console.alibabacloud.comdocs.modelstudio.console.alibabacloud.com/zh/model-studio/anthropic-api-messages
- docs.modelstudio.console.alibabacloud.comdocs.modelstudio.console.alibabacloud.com/zh/model-studio/qwen-structured-output
- docs.modelstudio.console.alibabacloud.comdocs.modelstudio.console.alibabacloud.com/zh/model-studio/list-models
完整 Token 价格
单位:美元 / 百万 Token。— 表示未知价格。上下文区间包含缓存读取和写入 Token,区间为左开右闭 (起点, 终点]。
| 计费项目 | 本站销售价 | 官方参考价 |
|---|---|---|
| 输入 | $1.5 | $1.5 |
| 缓存读取 | $0.075 | — |
| 缓存写入 · 5 min | $1.875 | — |
| 缓存写入 · 1 h | $3 | — |
| 输出 | $4.5 | $4.5 |
fc30311f41dcVendor-documented context window · 1,000,000 tokens · Maximum output · 262,144 tokens ·
Cost calculator
USD · EstimateAdvanced token usage
Input includes cache reads and writes; these portions must not overlap. Output already includes reasoning; enter reasoning as a subset, never add it twice. Count agent steps and retries as requests.
Estimated monthly cost
—USDQwen4-Max
Cache hits and capacity are not guaranteed.
Cost breakdown
- Uncached input
- —
- Cache read
- —
- Cache writes (5m + 1h)
- —
- Output (including reasoning)
- —
Usage limits
- Structured JSON output requires text input; image input cannot be combined with JSON format constraints.
- Video, audio, long-context quality, concurrent throughput and cross-request conversation storage were not verified.
- Configured limits do not establish full-window quality; input and actual output remain subject to the supplier context budget.
- Cache prices apply to actual supplier usage. Explicit cache creation/control was not verified.
- Cancellation without final usage retains a reservation for reconciliation. The reservation ceiling is not an estimated debit and dispatched requests are not automatically replayed.
Use cases
Reasoning and tool workflows
Image understanding
Structured data output
Integration tools and guides
FAQ
How do I call Qwen4-Max?
Use Model ID qwen/qwen4-max and the native fields in this page's API examples. Declared endpoints: /v1/chat/completions, /v1/responses, /v1/messages.
How is usage priced?
Use the current prices and billing units. Missing prices remain unknown, not free. The calculator uses native Gateway quotes; an estimate is not a bill.
Which parameters and limits apply?
Check the current parameter table and usage limits. Budget fields differ by protocol. Configured limits do not establish full-window quality or guarantee a maximum-length output.
New arrivals
Explore the models selected for New Arrivals, with API pricing, supported capabilities and integration guides.
Gemini 4 Hydrogen
gemini/gemini-4-hydrogen
Google Cloud lists Gemini 4 Hydrogen for fast coding assistance, multimodal input and high-volume text generation.
Launched on IteraCompute
- Input
- $1.50
- Output
- $7.50
USD / 1M tokens
GPT-6.2 Astra
openai/gpt-6.2-astra
OpenAI lists GPT-6.2 Astra for complex engineering and long-horizon reasoning.
Launched on IteraCompute
- Input
- $8.00
- Output
- $40.00
USD / 1M tokens
GPT-6.2 Sol
openai/gpt-6.2-sol
OpenAI lists GPT-6.2 Sol for everyday coding and agent workflows.
Launched on IteraCompute
- Input
- $2.00
- Output
- $10.00
USD / 1M tokens