Qwen3.8 Open-Weight and Qwen Image 3 Models Added

Date: 2026-08-15

Summary

AvalAI now provides three additional Alibaba models: the text-only, mandatory-thinking qwen3.8-2.4t-a95b open-weight model and the qwen-image-3.0-pro and qwen-image-3.0 image generation and editing models. The text model supports Chat Completions and Messages fully, with partial Responses support. Both image models support the Images generation and edit endpoints.

Details

Alibaba

qwen3.8-2.4t-a95b

qwen3.8-2.4t-a95b is the open-weight base underlying the managed qwen3.8-max service. Its mixture-of-experts architecture has 2.4 trillion total parameters and activates 95 billion parameters for each forward pass.

Unlike qwen3.8-max, the open-weight route is text-only and always operates in thinking mode. Thinking cannot be disabled. Use reasoning_effort with low, medium, or xhigh to tune the reasoning budget; xhigh is the default.

Feature	Details
Model ID	qwen3.8-2.4t-a95b
Input	Text
Output	Text
Context window	262,144 tokens
Architecture	2.4T total parameters, 95B active
Thinking	Mandatory
Endpoints	v1/chat/completions (full), v1/messages (full), v1/responses (partial)

Pricing:

Usage	Price per 1M tokens
Input	$2.00
Cache creation input	$2.50
Cached input	$0.25
Output	$6.00

qwen-image-3.0-pro and qwen-image-3.0

Both Qwen Image 3 models support new-image generation and source-image editing through the OpenAI-compatible Images API.

Model ID	Generation endpoint	Edit endpoint
qwen-image-3.0-pro	v1/images/generations	v1/images/edits
qwen-image-3.0	v1/images/generations	v1/images/edits

Both models use the same pricing:

Usage	Price
Output accounting rate	$40 / 1M output tokens
1K / approximately 1 MP output image	$0.04 / image
2–4 MP output image	$0.075 / image
Reference or input image	$0.003 / image
Text input	$0
Cached text input	$0

API Request and Response Examples

Chat Completions

Example Request

curl https://api.avalai.ir/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $AVALAI_API_KEY" \
  -d '{
    "model": "qwen3.8-2.4t-a95b",
    "messages": [
      {
        "role": "user",
        "content": "Identify the highest-risk assumption in this migration plan."
      }
    ],
    "reasoning_effort": "medium"
  }'

Example Response

The model's text can begin with a <think>...</think> reasoning block because thinking is mandatory.

{
  "id": "chatcmpl_qwen38_example",
  "object": "chat.completion",
  "model": "qwen3.8-2.4t-a95b",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "<think>Reviewing dependency and rollback assumptions...</think>\n\nThe highest-risk assumption is that every dependent service can migrate in the same maintenance window. Validate compatibility service by service and define a reversible cutover."
      },
      "finish_reason": "stop"
    }
  ]
}

Image Generation

Example Request

curl https://api.avalai.ir/v1/images/generations \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $AVALAI_API_KEY" \
  -d '{
    "model": "qwen-image-3.0-pro",
    "prompt": "A clean editorial illustration of a renewable-energy city, with room for a headline",
    "size": "1024x1024",
    "n": 1
  }'

Example Response

{
  "created": 1786795200,
  "data": [
    {
      "url": "https://example.invalid/generated/qwen-image-3-output.png",

      "revised_prompt": "A clean editorial illustration of a renewable-energy city, with clear headline space"
    }
  ]
}

For editing, submit a source image and prompt to v1/images/edits. Each reference or input image adds $0.003 to the request cost.

SDK Usage Examples

Bash

curl https://api.avalai.ir/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $AVALAI_API_KEY" \
  -d '{
    "model": "qwen3.8-2.4t-a95b",
    "messages": [{"role": "user", "content": "Review this deployment plan."}],
    "reasoning_effort": "medium"
  }'

Python

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["AVALAI_API_KEY"],
    base_url="https://api.avalai.ir/v1",
)

response = client.chat.completions.create(
    model="qwen3.8-2.4t-a95b",
    messages=[{"role": "user", "content": "Review this deployment plan."}],
    extra_body={"reasoning_effort": "medium"},
)

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

JavaScript

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.AVALAI_API_KEY,
  baseURL: "https://api.avalai.ir/v1",
});

const response = await client.chat.completions.create({
  model: "qwen3.8-2.4t-a95b",
  messages: [
    { role: "user", content: "Review this deployment plan." },
  ],
  reasoning_effort: "medium",
});

console.log(response.choices[0].message.content);

Integration Notes

• Do not send image or video input to qwen3.8-2.4t-a95b; use qwen3.8-max when multimodal understanding is required.
• Do not attempt to disable thinking on qwen3.8-2.4t-a95b.
• Use reasoning_effort rather than the managed Max model's enable_thinking control.
• Treat v1/responses support for the text model as partial and verify required fields before migration.
• Use the plural edit path, v1/images/edits, for both Qwen Image 3 models.
• Account for the $0.003 fee for every reference/input image in editing workflows.

Related Links

• Alibaba model documentation
• Qwen3.8-2.4T-A95B model page
• Qwen Image 3.0 Pro model page
• Qwen Image 3.0 model page
• Chat Completions API
• Images API
• Reasoning guide
• Image generation guide
• Pricing