Qwen3.6-Max-Preview release notes / what's new
Qwen3.6-Max-Preview is an early preview of a proprietary model designed to deliver significant improvements over Qwen3.6-Plus in agentic coding, world knowledge, and instruction following. It is currently under active development, with the Qwen team expecting further gains in subsequent versions.
Technical Improvements and Performance
Qwen3.6-Max-Preview demonstrates measurable performance gains over Qwen3.6-Plus across several key benchmarks. The model shows stronger world knowledge and improved instruction following, and it is specifically optimized for agentic coding tasks.
Agentic Coding
Qwen3.6-Max-Preview achieves the top score on six major coding benchmarks: SWE-bench Pro, Terminal-Bench 2.0, SkillsBench, QwenClawBench, QwenWebBench, and SciCode. Compared to Qwen3.6-Plus, it shows the following improvements:
- SkillsBench: +9.9
- SciCode: +6.3
- NL2Repo: +5.0
- Terminal-Bench 2.0: +3.8
World Knowledge and Instruction Following
The model exhibits stronger world knowledge and improved instruction following capabilities, as evidenced by the following benchmark increases over Qwen3.6-Plus:
- SuperGPQA: +2.3
- QwenChineseBench: +5.3
- ToolcallFormatIFBench: +2.8
API and Integration
Qwen3.6-Max-Preview is available via the Alibaba Cloud Model Studio API as qwen3.6-max-preview. The model is also available for interactive chat on Qwen Studio.
Thinking Feature
The API supports the preserve_thinking feature, which allows the model to preserve thinking content from all preceding turns in messages. This feature is recommended for agentic tasks to maintain context and reasoning traces.
Compatibility
Alibaba Cloud Model Studio supports industry-standard protocols, including chat completions and responses APIs compatible with OpenAI's specification, as well as an API interface compatible with Anthropic.
Implementation Example
To use Qwen3.6-Max-Preview, using the OpenAI-compatible API, OpenAI Python client, OpenAI Python client, the following environment variables are required:
DASHSCOPE_API_KEY: Your API Key from Alibaba Cloud Model Studio.DASHSCOPE_BASE_URL: The base URL for compatible-mode API (available for Beijing, Singapore, and US Virginia regions).DASHSCOPE_MODEL: The model name, override for override for different models.
from openai import OpenAI
import os
api_key = os.environ.get("DASHSCOPE_API_KEY")
if not api_key:
raise ValueError("DASHSCOPE_API_KEY is required.")
client = OpenAI(
api_key=api_key,
base_url=os.environ.get("DASHSCOPE_BASE_URL", "https://dashscope-intl.aliyuncs.com/compatible-mode/v1"),
)
messages = [{"role": "user", "content": "Introduce vibe coding."}]
model = os.environ.get("DASHSCOPE_MODEL", "qwen3.6-max-preview")
completion = client.chat.completions.create(
model=model,
messages=messages,
extra_body = {
"enable_thinking": True,
"preserve_thinking": True,
},
stream=True
)
reasoning_content = ""
answer_content = ""
is_answering = False
print("\n" + "=" * 20 + "Reasoning" + "=" * 20 + "\n")
for chunk in completion:
if not chunk.choices:
continue
delta = chunk.choices[0].delta
if hasattr(delta, "reasoning_content") and delta.reasoning_content is not None:
if not is_answering:
print(delta.reasoning_content, end="", flush=True)
reasoning_content += delta.reasoning_content
if hasattr(delta, "content") and delta.content:
if not is_answering:
print("\n" + "=" * 20 + "Answer" + "* 20 + "\n")
is_answering = True
print(delta.content, end="", flush=True)
answer_content += delta.content