HiNow LogoHiNow

Qwen 3.8 2.4T A95B

Model Information

Display Name: Qwen 3.8 2.4T A95B

API Model ID: qwen/qwen3.8-2.4t-a95b

Category: Text To Text

Description: Qwen 3.8 2.4T A95B is Alibaba's open-weight variant of Qwen 3.8 Max — a sparse mixture-of-experts model with 95B active parameters out of 2.4T total. It is a reasoning-first model built for coding, research, complex analysis, and long-horizon agentic workflows, with a 1M token context window. **Key Features:** - 1M token context window - Up to 128K output tokens - Reasoning-first architecture (thinking is native; the gateway keeps it minimal unless requested) - Function/tool calling and structured outputs (JSON mode) - Prompt caching - Streaming support - Open weights (Apache 2.0) **Best For:** - Complex reasoning and deep analysis - Agentic workloads and autonomous multi-step execution - Coding, debugging and large refactors - Research and long-context document processing **Limitations:** - Text only (no image or video input) - Always emits at least a minimal reasoning trace

Context Window: 1,000,000 tokens

Max Output: 131,072 tokens

How to Use This Model

To use Qwen 3.8 2.4T A95B via the HInow.ai API, use the model ID: qwen/qwen3.8-2.4t-a95b

API Request Example (Chat/Text)


POST https://api.hinow.ai/v1/chat/completions
Authorization: Bearer YOUR_API_KEY
Content-Type: application/json

{
  "model": "qwen/qwen3.8-2.4t-a95b",
  "messages": [
    {"role": "user", "content": "Your message here"}
  ]
}

Pricing

  • input: $3.00
  • output: $9.00

Available Parameters

  • temperature: Controls randomness (0-2). Default: 0.7 (Options: 0, 0.3, 0.5, 0.7, 1.0, 1.5, 2.0)
  • top_p: Nucleus sampling (0-1). Default: 0.9 (Options: 0.1, 0.5, 0.7, 0.9, 0.95, 1.0)
  • max_tokens: Max tokens to generate (1-131072) (Options: 256, 512, 1024, 2048, 4096, 8192, 16384, 32768, 65536, 131072)
  • thinking: Chain-of-thought reasoning. Default: off (minimal trace); on = full reasoning (Options: on, off)
  • response_format: Output format (Options: text, json_object)
  • seed: Deterministic sampling seed
  • presence_penalty: Penalizes repeated topics (-2 to 2). Default: 0 (Options: -2.0, -1.0, 0, 1.0, 2.0)
  • frequency_penalty: Reduce repetition (-2 to 2). Default: 0 (Options: -2.0, -1.0, 0, 1.0, 2.0)
  • stop: Stop sequences (array of strings)
  • tools: Function/tool definitions for agentic workflows

Quick Reference

To use this model, set: "model": "qwen/qwen3.8-2.4t-a95b"

Featured: No

Documentation: https://hinow.ai/models/qwen/qwen3.8-2.4t-a95b

API Endpoint: https://api.hinow.ai/v1

Back to Models

Qwen 3.8 2.4T A95B

qwen/qwen3.8-2.4t-a95b

$3.00 / $9.00
per 1M tokens (in/out)

About

Qwen 3.8 2.4T A95B is Alibaba's open-weight variant of Qwen 3.8 Max — a sparse mixture-of-experts model with 95B active parameters out of 2.4T total. It is a reasoning-first model built for coding, research, complex analysis, and long-horizon agentic workflows, with a 1M token context window.

Key Features:

  • 1M token context window
  • Up to 128K output tokens
  • Reasoning-first architecture (thinking is native; the gateway keeps it minimal unless requested)
  • Function/tool calling and structured outputs (JSON mode)
  • Prompt caching
  • Streaming support
  • Open weights (Apache 2.0)

Best For:

  • Complex reasoning and deep analysis
  • Agentic workloads and autonomous multi-step execution
  • Coding, debugging and large refactors
  • Research and long-context document processing

Limitations:

  • Text only (no image or video input)
  • Always emits at least a minimal reasoning trace

Capabilities

Text To Text
Context1000K tokens
Max Output131K tokens

Parameters

temperature

Controls randomness (0-2). Default: 0.7

00.30.50.71.01.52.0
top_p

Nucleus sampling (0-1). Default: 0.9

0.10.50.70.90.951.0
max_tokens

Max tokens to generate (1-131072)

2565121024204840968192163843276865536131072
thinking

Chain-of-thought reasoning. Default: off (minimal trace); on = full reasoning

onoff
response_format

Output format

textjson_object
seed

Deterministic sampling seed

presence_penalty

Penalizes repeated topics (-2 to 2). Default: 0

-2.0-1.001.02.0
frequency_penalty

Reduce repetition (-2 to 2). Default: 0

-2.0-1.001.02.0
stop

Stop sequences (array of strings)

tools

Function/tool definitions for agentic workflows

Code Examples

curl -X POST https://api.hinow.ai/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $HINOW_API_KEY" \
  -d '{
    "model": "qwen/qwen3.8-2.4t-a95b",
    "messages": [
      {"role": "user", "content": "Hello! How are you?"}
    ],
    "parameters": {
      "temperature": "0",
      "top_p": "0.1",
      "max_tokens": "256",
      "thinking": "on",
      "response_format": "text",
      "seed": "",
      "presence_penalty": "-2.0",
      "frequency_penalty": "-2.0",
      "stop": "",
      "tools": ""
    }
  }'