DashScope Qwen3.7 Series Complete Guide: Max, Plus, and Multimodal API Reference

The Qwen3.7 series is Alibaba's current flagship model generation on DashScope. It ships two models that split along a clear axis: Qwen3.7-Max handles text-only reasoning and agentic coding, while Qwen3.7-Plus adds vision and video input at roughly one-third the price. Both share a 1M-token context window, support thinking mode, and expose an OpenAI-compatible API endpoint — meaning you can integrate them with a three-line config change in any OpenAI SDK client.

We wrote this guide because the Qwen3.7 series replaces several older models sunsetting in September 2026. If you are on qwen3-max, qwen3-coder-plus, or any Qwen3.5/3.6 variant, this is where you land. For the migration checklist, see our September 2026 Qwen Model Sunset Migration Guide.

Qwen3.7 Series at a Glance

Source: Aliyun Model Studio Pricing, Alibaba Cloud Model Studio Models. Retrieved 2026-06-24.

Qwen3.7-Max — Text Reasoning Flagship

Qwen3.7-Max is Alibaba's top reasoning model. It scored 80.4% on SWE-Bench Verified and is positioned as an agent-first model optimized for coding and complex multi-step tasks. It was released on May 20, 2026, with an updated snapshot on June 8, 2026.

Key capabilities:

• Extended-thinking mode for multi-step planning, code refactoring, and long agent chains
• 1M-token context for large codebases and document analysis
• Tool calling with tools and tool_choice parameters
• Structured output via response_format
• Batch API support at 50% off standard pricing

Model IDs available on DashScope:

• qwen3.7-max — rolling pointer to latest stable (currently qwen3.7-max-2026-05-20)
• qwen3.7-max-2026-06-08 — June 2026 snapshot with multimodal upgrade
• qwen3.7-max-2026-05-20 — initial release snapshot
• qwen3.7-max-preview — preview channel (thinking-mode only)

Source: Qwen3.7: The Agent Frontier, MarkTechPost coverage. Retrieved 2026-06-24.

Qwen3.7-Plus — Multimodal Agent Intelligence

Qwen3.7-Plus handles text, image, and video input at roughly one-third the cost of Qwen3.7-Max. It was announced on May 31, 2026, and is designed for multimodal agent workflows — generating interactive webpages from visual references, understanding video materials, and processing design intent.

Key capabilities:

• Vision input: images in base64 or URL format
• Video input: video file understanding
• Thinking mode support (same as Max)
• Agentic coding with tool calling
• 1M-token context shared across all modalities

Model ID: qwen3.7-plus

When to pick Plus over Max: use Qwen3.7-Plus when your workload involves image or video understanding, or when you need strong text performance at a lower cost and do not require the absolute highest reasoning capability.

Source: Qwen3.7-Plus: Multimodal Agent Intelligence, VentureBeat coverage. Retrieved 2026-06-24.

Getting Started — API Key and Endpoint Setup

Step 1: Create your DashScope account. Go to Alibaba Cloud Model Studio (international) or 百炼控制台 (China mainland).

Step 2: Get an API key. Navigate to API Keys in the console and create one. You can scope keys to specific workspaces for multi-team isolation.

Step 3: Install the OpenAI SDK:

pip install --upgrade openai

Step 4: Make your first call to Qwen3.7-Max:

from openai import OpenAI
import os

client = OpenAI(
    api_key=os.getenv("DASHSCOPE_API_KEY"),
    base_url="https://dashscope.aliyuncs.com/compatible-mode/v1",
)

response = client.chat.completions.create(
    model="qwen3.7-max",
    messages=[
        {"role": "system", "content": "You are a senior software engineer."},
        {"role": "user", "content": "Write a Python function that implements exponential backoff with jitter."},
    ],
    max_tokens=4096,
)

print(response.choices[0].message.content)

For the international endpoint, use https://dashscope-intl.aliyuncs.com/compatible-mode/v1 instead.

Source: Alibaba Cloud Model Studio — OpenAI Compatibility. Retrieved 2026-06-24.

Thinking Mode Configuration

Both Qwen3.7-Max and Qwen3.7-Plus support thinking mode — the model reasons step by step before delivering a final answer. This is useful for complex problems in math, coding, and logic.

To enable thinking mode, set the enable_thinking parameter to true in extra body:

response = client.chat.completions.create(
    model="qwen3.7-max",
    messages=[
        {"role": "user", "content": "Prove that there are infinitely many primes."},
    ],
    max_tokens=16384,
    extra_body={
        "enable_thinking": True,
        "thinking_budget": 10000,  # optional: cap thinking tokens
    },
)

# The thinking content appears in the response
for choice in response.choices:
    if hasattr(choice.message, "reasoning_content"):
        print("Thinking:", choice.message.reasoning_content)
    print("Answer:", choice.message.content)

Key considerations:

• Thinking tokens count toward output pricing (¥36/1M for Max, ¥12.32/1M for Plus)
• thinking_budget caps how many tokens the model spends on reasoning
• Non-thinking mode (default) is faster and cheaper for straightforward tasks
• qwen3.7-max-preview only supports thinking mode — it cannot run in non-thinking mode

Source: Qwen3: Think Deeper, Act Faster, Aliyun Model Studio Pricing. Retrieved 2026-06-24.

Multimodal Input — Vision and Video (Qwen3.7-Plus)

Qwen3.7-Plus accepts images and video as input alongside text. Here is how to send an image for visual understanding:

response = client.chat.completions.create(
    model="qwen3.7-plus",
    messages=[
        {
            "role": "user",
            "content": [
                {"type": "text", "text": "What does this architecture diagram show?"},
                {
                    "type": "image_url",
                    "image_url": {
                        "url": "https://example.com/architecture-diagram.png",
                    },
                },
            ],
        }
    ],
    max_tokens=2048,
)

print(response.choices[0].message.content)

You can also send base64-encoded images:

import base64

with open("diagram.png", "rb") as f:
    image_data = base64.b64encode(f.read()).decode()

response = client.chat.completions.create(
    model="qwen3.7-plus",
    messages=[
        {
            "role": "user",
            "content": [
                {"type": "text", "text": "Describe this image."},
                {
                    "type": "image_url",
                    "image_url": {
                        "url": f"data:image/png;base64,{image_data}",
                    },
                },
            ],
        }
    ],
)

For video input, DashScope supports passing a video URL in the same format. Check the DashScope multimodal documentation for supported video formats and size limits.

Source: Alibaba Cloud Model Studio Models. Retrieved 2026-06-24.

Pricing and Rate Limits

Pricing Breakdown (China Mainland — Beijing Region)

International Pricing (Singapore Region)

Singapore pricing is roughly 1.56× the China mainland rate.

Cost-Saving Options

• Batch API: 50% off both input and output for Qwen3.7-Max. Available for non-real-time workloads.
• Context cache: Repeated context segments (system prompts, few-shot examples) are cached and billed at a discount. See the next section.

Comparison vs Predecessors

Qwen3.7-Max has a flat rate across the entire 1M context window — no tiered pricing — which simplifies cost estimation for long-context workloads.

Source: Aliyun Model Studio Pricing. Retrieved 2026-06-24.

Context Cache and Cost Optimization

DashScope supports two cache modes for Qwen3.7 models:

1. Implicit cache: DashScope automatically detects repeated context prefixes across requests. No configuration needed — cached tokens are billed at a discount automatically.
2. Explicit cache: You manually pin specific content segments (system prompts, reference documents) to guarantee cache hits. Useful for agent workflows where the system prompt is large and stable.

To use explicit cache:

response = client.chat.completions.create(
    model="qwen3.7-max",
    messages=[
        {
            "role": "system",
            "content": "You are a code reviewer for a Python monorepo...",  # large prompt
        },
        {"role": "user", "content": "Review this diff: ..."},
    ],
    extra_body={
        "enable_context_cache": True,
    },
)

# Check cache usage in response
print(response.usage)  # includes cached_tokens field

For workloads with repeated system prompts, context cache can reduce effective input costs by 60–80%. For detailed configuration, see the DashScope Context Cache documentation.

Source: Alibaba Cloud Model Studio — Context Cache, Explicit Cache Best Practices. Retrieved 2026-06-24.

Routing Qwen3.7 Models via TheRouter

We route both Qwen3.7-Max and Qwen3.7-Plus through Aliyun Bailian on TheRouter. The model IDs in your routing configuration are:

• qwen/qwen3.7-max — text reasoning flagship
• qwen/qwen3.7-plus — multimodal, cost-effective

You can set up fallback routing so that if Qwen3.7-Max is unavailable, your traffic falls back to Qwen3.7-Plus (or another provider). For implementation details, see our LLM API Fallback Routing Guide.

An example routing configuration:

routes:
  - model: qwen/qwen3.7-max
    provider: aliyun-bailian
    fallback:
      - model: qwen/qwen3.7-plus
        provider: aliyun-bailian
      - model: deepseek/deepseek-v4-pro
        provider: deepseek

This routes text-heavy workloads through Qwen3.7-Max by default, falling back to Plus (same provider, lower cost) and then DeepSeek V4 (different provider) if needed.

Migration from Qwen3/3.5/3.6 Series

If you are currently on older Qwen models, here is the migration path:

Critical dates: qwen3-max (snapshot 2025-09-23) and qwen3-coder-plus are sunsetting September 8, 2026. After that date, requests to those model IDs will fail. See our September 2026 Qwen Model Sunset Migration Guide for the complete timeline.

The migration is a model ID swap — the API endpoint and authentication remain identical:

# Before
response = client.chat.completions.create(model="qwen3-max", ...)

# After
response = client.chat.completions.create(model="qwen3.7-max", ...)

Common Errors and Fixes

FAQ

Q: Is Qwen3.7-Max the same as Qwen3-Max? No. Qwen3.7-Max is the successor with a 1M-token context window (up from 256K), improved reasoning, and flat pricing. Qwen3-Max is being deprecated on September 8, 2026.

Q: Can I use Qwen3.7-Max for vision tasks? No. Qwen3.7-Max is text-only. Use Qwen3.7-Plus for image and video input.

Q: Are thinking tokens billed separately? No. Thinking tokens are billed at the same output rate. The thinking_budget parameter lets you control how many tokens the model uses for reasoning.

Q: Is Qwen3.7-Max open source? No. Both Qwen3.7-Max and Qwen3.7-Plus are proprietary models available only through DashScope API.

Q: Does the free tier apply to Qwen3.7 models? Yes. New Alibaba Cloud accounts get 1M free tokens per model, valid for 90 days after activating Model Studio (China mainland only).

---

Last updated: 2026-06-24. Pricing and model availability verified against Aliyun Model Studio and Alibaba Cloud Model Studio.