Qwen3.8-2.4T-A95B

Flagship-scale model for coding and long-horizon agentic work with 256K context

Try now

read docs

About model

Qwen3.8 Max is the flagship of Qwen's 3.8 generation and the largest model the team has released, a sparse Mixture-of-Experts design with 2.4 trillion total parameters. Qwen describes it as the team's first multimodal model above one trillion parameters, built for coding, long-horizon agentic work, and multimodal reasoning. The model supports a 256K-token context window with up to 128K output tokens, and thinking is always enabled, with reasoning effort adjustable across low, high, and xhigh levels. Announced on July 19, 2026 at the World AI Conference in Shanghai, with Alibaba committing to an open-weight release. Available on Together AI.

‍

Total Parameters

2.4T

Sparse MoE design, the largest model Qwen has released

Context Window

256K

Entire repositories or thousands of pages, with up to 128K output tokens

Reasoning Effort Levels

Always-on thinking, adjustable across low, high, and xhigh per request

Model key capabilities

Coding & Agentic Work: Positioned by Qwen around coding and long-horizon agentic tasks, with reasoning always enabled
Multimodal Input: Processes text and images together, Qwen's first multimodal model at this scale
Adjustable Reasoning Effort: Low, high, and xhigh thinking levels set per request, trading response speed against deliberation depth
Production-Ready Infrastructure: 99.9% SLA, available on serverless and dedicated infrastructure

API usage

cURL
Python
Typescript

Endpoint:

Qwen/Qwen3.8-Max

curl -X POST "https://api.together.xyz/v1/chat/completions" \
  -H "Authorization: Bearer $TOGETHER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "Qwen/Qwen3.8-Max",
    "messages": [
      {
        "role": "user",
        "content": "What are some fun things to do in New York?"
      }
    ]
}'

from together import Together

client = Together()

response = client.chat.completions.create(
  model="Qwen/Qwen3.8-Max",
  messages=[
    {
      "role": "user",
      "content": "What are some fun things to do in New York?"
    }
  ]
)
print(response.choices[0].message.content)

import Together from 'together-ai';
const together = new Together();

const completion = await together.chat.completions.create({
  model: 'Qwen/Qwen3.8-Max',
  messages: [
    {
      role: 'user',
      content: 'What are some fun things to do in New York?'
     }
  ],
});

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

Model card
Architecture Overview:
• Sparse Mixture-of-Experts model with 2.4 trillion total parameters
• 256K-token context window with a 131,072-token maximum output
• Thinking always enabled, with reasoning effort configurable across low, high, and xhigh levels (xhigh default)
• Multimodal input processed natively alongside text

Training Methodology:
• Announced July 19, 2026 at the World AI Conference in Shanghai as the flagship of the Qwen3.8 generation
• Qwen describes it as the team's first multimodal model above one trillion parameters
• The preview period saw rapid iteration, with Qwen-reported gains concentrated in web frontend generation

Performance Characteristics:
• Qwen positions the model around coding, long-horizon agentic work, and multimodal reasoning
• Qwen-reported improvements to web frontend generation during the July preview updates
‍
Prompting
Together AI API Access:
• Access Qwen3.8 Max via Together AI APIs using the endpoint Qwen/Qwen3.8-Max
• Authenticate using your Together AI API key in request headers
• Thinking is always enabled; set reasoning effort to low, high, or xhigh per request (xhigh is the default)
• Supports output lengths up to 128K tokens for long generations and agent trajectories
• Available on Together AI serverless and dedicated infrastructure
‍
Applications & use cases
Agentic Software Development:
• Run long-horizon coding sessions with always-on reasoning tuned to task difficulty
• Generate frontend and full-stack work, an area of Qwen-reported preview gains
• Hold entire repositories in the 256K-token context for cross-file work

Long-Horizon Agent Workflows:
• Sustain extended multi-step agent trajectories with up to 128K output tokens per generation
• Dial reasoning effort down for routine steps and up to xhigh for hard planning
• Keep long-running agent state in context across sessions

Multimodal Reasoning:
• Reason over documents, screenshots, and images alongside text in one request
• Ground coding and analysis tasks in visual context without a separate pipeline
• Combine multimodal input with long context for large mixed-media working sets
‍

Related models

Model specifications

Model data

Model provider
Qwen
Type
Chat
Reasoning
Vision
Main use cases
Coding Agents
Features
Function Calling
JSON Mode
Deployment
Serverless
Endpoint
Qwen/Qwen3.8-Max
Parameters
2.4T
Context length
256K
Input price
$2.50 / 1M tokens
$0.50 (cached)/1M
Output price
$6.25 / 1M tokens
Input modalities
Text
Image
Output modalities
Text

Released
August 2, 2026
Category
Chat

Run in Playground

Quickstart docs

Deploy model

Qwen3.8-2.4T-A95B

About model

API usage

Model card

Prompting

Applications & use cases