Your privacy choices

Allow optional cookies for referral attribution, visit analytics, and Google Ads purchase measurement.

Back to blog
中文English

Qwen 3.8 Max API Pricing & Setup Guide (2026)

Qwen 3.8 MaxQwen APIAlibaba Qwen1M contextAI agents

The qwen3.8-max route is now available through this gateway. As of July 20, 2026, the platform catalog lists it at a 10x multiplier (CNY 4.00 per 1M tokens), with a 1M context window and vision support. It is positioned for complex reasoning, coding, long-running agents, research, long documents, and multimodal work.

There is an important naming boundary. qwen3.8-max is this gateway's route ID. Qwen's public Qwen3-Max article names its official API model qwen3-max. Similar names are not proof that two routes are identical, and you should not substitute one ID for the other. This guide documents platform access, pricing, and validation; it is not an Alibaba Cloud launch announcement.

Qwen 3.8 Max API at a glance

Item Current gateway listing
Model ID qwen3.8-max
Rate 10x
Reference price CNY 4.00 per 1M tokens
Context window 1M
Vision Supported
Best first tests Reasoning, coding, agents, long documents, image understanding
API format OpenAI-compatible Chat Completions

These are gateway catalog settings. Recheck the live pricing page and model guide before production use because routes, multipliers, and capability labels can change.

How to call qwen3.8-max

Use the regional Base URL recommended in the integration guide. The following minimal request uses the global endpoint:

curl https://gpt-agent.cc/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen3.8-max",
    "messages": [
      {"role": "user", "content": "Create a three-step refactoring plan and list the main risks."}
    ]
  }'

With the OpenAI Python SDK, replace the Base URL and model ID while keeping the standard Chat Completions request shape:

from openai import OpenAI

client = OpenAI(
    api_key="YOUR_API_KEY",
    base_url="https://gpt-agent.cc/v1",
)

response = client.chat.completions.create(
    model="qwen3.8-max",
    messages=[
        {"role": "user", "content": "Turn this requirement into an implementation plan and acceptance criteria."}
    ],
)

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

Start with a short text request to validate authentication, the exact model ID, and response parsing. Test tool calls, long context, and image inputs separately before moving production traffic.

How the gateway price works

The platform baseline is approximately CNY 0.40 per 1M tokens at 1x. At the current 10x rate, the qwen3.8-max reference price is CNY 4.00 per 1M tokens.

Estimated usage = billed tokens / 1,000,000 * CNY 4.00

For example, 250,000 billed tokens would be approximately CNY 1.00 under this simplified estimate. The actual account ledger, cache treatment, and current pricing page remain authoritative. For long-document workloads, avoid resending unchanged context on every turn when your application can reuse or summarize it safely.

Qwen 3.8 Max vs Kimi K3 on this gateway

The gateway currently lists both qwen3.8-max and kimi-k3 with 1M context and vision support. Their current rates and source boundaries differ:

Gateway route Current rate Good first evaluation set
qwen3.8-max 10x Cost-sensitive reasoning, coding, agent, and multimodal tasks
kimi-k3 15x Long-horizon coding, knowledge work, agents, and K3-specific evaluations

This is not a benchmark ranking. Build a representative set of 20 to 50 tasks and compare completion rate, latency, human rework, and total cost under the same tools and acceptance criteria. For source-backed Kimi specifications, see the Kimi K3 API pricing and specs guide.

Production verification checklist

  1. Copy qwen3.8-max from the model guide; do not silently replace it with qwen3-max.
  2. Run a short text request and record status code, request ID, latency, and output.
  3. Test image input using the exact client message format you will deploy.
  4. Increase context size gradually and measure time to first token, total latency, and billed usage.
  5. For agents, validate tool schemas, retries, timeouts, and idempotency.
  6. Recheck pricing and keep a lower-cost fallback route before launch.

Frequently asked questions

What is the Qwen 3.8 Max model ID?

Use qwen3.8-max on this gateway. Qwen's public article uses qwen3-max for its official API model; the IDs are not interchangeable.

Does Qwen 3.8 Max support 1M context?

The current gateway catalog lists 1M. Effective request size can still be reduced by the client, message format, reserved output, or gateway limits, so test progressively.

Does Qwen 3.8 Max accept images?

The route is currently marked as supporting image understanding. Image URLs, Base64 payloads, and multimodal message structures vary across clients, so verify one real request with the client you plan to use.

Where can I buy access?

New users can obtain a multi-model key from the buy page. Existing key holders should use the top-up page. This is an independent compatible gateway, not an official Alibaba Cloud or Qwen storefront.

Sources and next steps