Qwen3.6-35B-A3B 开源：35B 总参数 3B 激活的 MoE 模型

Qwen3.6-35B-A3B: Agentic Coding Power, Now Open to All | Qwen Blog Publication About Try Qwen Chat Qwen3.6-35B-A3B: Agentic Coding Power, Now Open to All April 15, 2026 · 8 min · 1639 words · Qwen Team | Translations: 简体中文 QWEN STUDIO HUGGING FACE MODELSCOPE DISCORD Following the launch of Qwen3.6-Plus , we are excited to open-source Qwen3.6-35B-A3B — a sparse yet remarkably capable mixture-of-experts (MoE) model with 35 billion total parameters and only 3 billion active parameters. Despite its efficiency, Qwen3.6-35B-A3B delivers outstanding agentic coding performance, surpassing its predecessor Qwen3.5-35B-A3B by a wide margin and rivaling much larger dense models such as Qwen3.5-27B and Gemma4-31B. Still supporting both multimodal thinking and non-thinking modes, Qwen3.6-35B-A3B works as one of the most versatile open-source models available today. Now, Qwen3.6-35B-A3B is live on Qwen Studio, available through our API, and released as open weights for the community. Qwen3.6-35B-A3B is a fully open-source MoE model (35B total / 3B active), featuring: exceptional agentic coding capability competitive with much larger models strong multimodal perception and reasoning ability You can chat interactively on Qwen Studio , call via API as Qwen3.6-Flash on Alibaba Cloud Model Studio API , or download weights from Hugging Face and ModelScope . Performance # Below we present comprehensive evaluations of Qwen3.6-35B-A3B against peer-scale models across a wide range of tasks and modalities. Language # With only 3B active parameters, Qwen3.6-35B-A3B outperforms the dense 27B-parameter Qwen3.5-27B on several key coding benchmarks and dramatically surpasses its direct predecessor Qwen3.5-35B-A3B, especially on agentic coding and reasoning tasks. Qwen3.5-27B Gemma4-31B Qwen3.5-35BA3B Gemma4-26BA4B Qwen3.6-35BA3B Coding Agent SWE-bench Verified 75.0 52.0 70.0 17.4 73.4 SWE-bench Multilingual 69.3 51.7 60.3 17.3 67.2 SWE-bench Pro 51.2 35.7 44.6 13.8 49.5 Terminal-Bench 2.0 41.6 42.9 40.5 34.2 51.5 Claw-Eval Avg 64.3 48.5 65.4 58.8 68.7 Claw-Eval Pass^3 46.2 25.0 51.0 28.0 50.0 SkillsBench Avg5 27.2 23.6 4.4 12.3 28.7 QwenClawBench 52.2 41.7 47.7 38.7 52.6 NL2Repo 27.3 15.5 20.5 11.6 29.4 QwenWebBench 1068 1197 978 1178 1397 General Agent TAU3-Bench 68.4 67.5 68.9 59.0 67.2 VITA-Bench 41.8 43.0 29.1 36.9 35.6 DeepPlanning 22.6 24.0 22.8 16.2 25.9 Tool Decathlon 31.5 21.2 28.7 12.0 26.9 MCPMark 36.3 18.1 27.0 14.2 37.0 MCP-Atlas 68.4 57.2 62.4 50.0 62.8 WideSearch 66.4 35.2 59.1 38.3 60.1 Knowledge MMLU-Pro 86.1 85.2 85.3 82.6 85.2 MMLU-Redux 93.2 93.7 93.3 92.7 93.3 SuperGPQA 65.6 65.7 63.4 61.4 64.7 C-Eval 90.5 82.6 90.2 82.5 90.0 STEM & Reasoning GPQA 85.5 84.3 84.2 82.3 86.0 HLE 24.3 19.5 22.4 8.7 21.4 LiveCodeBench v6 80.7 80.0 74.6 77.1 80.4 HMMT Feb 25 92.0 88.7 89.0 91.7 90.7 HMMT Nov 25 89.8 87.5 89.2 87.5 89.1 HMMT Feb 26 84.3 77.2 78.7 79.0 83.6 IMOAnswerBench 79.9 74.5 76.8 74.3 78.9 AIME26 92.6 89.2 91.0 88.3 92.7 * SWE-Bench Series: Internal agent scaffold (bash + file-edit tools); temp=1.0, top_p=0.95, 200K context window. We correct some problematic tasks in the public set of SWE-bench Pro and evaluate all baselines on the refined benchmark. * Terminal-Bench 2.0: Harbor/Terminus-2 harness; 3h timeout, 32 CPU/48 GB RAM; temp=1.0, top_p=0.95, top_k=20, max_tokens=80K, 256K ctx; avg of 5 runs. * SkillsBench: Evaluated via OpenCode on 78 tasks (self-contained subset, excluding API-dependent tasks); avg of 5 runs. * NL2Repo: Others are evaluated via Claude Code (temp=1.0, top_p=0.95, max_turns=900). * QwenClawBench: An internal real-user-distribution Claw agent benchmark (open-sourcing soon); temp=0.6, 256K ctx. * QwenWebBench: An internal front-end code generation benchmark; bilingual (EN/CN), 7 categories (Web Design, Web Apps, Games, SVG, Data Visualization, Animation, and 3D); auto-render + multimodal judge (code/visual correctness); BT/Elo rating system. * TAU3-Bench: We use the official user model (gpt-5.2, low reasoning effort) + default BM25 retrieval. * VITA-Bench: Avg subdomain scores; using claude-4-sonnet as judger, as the official judger (claude-3.7-sonnet) is no longer available. * MCPMark: GitHub MCP v0.30.3; Playwright responses truncated at 32K tokens. * MCP-Atlas: Public set score; gemini-2.5-pro judger. * AIME 26: We use the full AIME 2026 (I & II), where the scores may differ from Qwen 3.5 notes. Vision Language # Qwen3.6 is natively multimodal, and Qwen3.6-35B-A3B showcases perception and multimodal reasoning capabilities that far exceed what its size would suggest, with only around 3 billion activated parameters. Across most vision-language benchmarks, its performance matches Claude Sonnet 4.5, and even surpasses it on several tasks. Its strengths are particularly evident in spatial intelligence, where it achieves 92.0 on RefCOCO and 50.8 on ODInW13. Qwen3.5-27B Claude-Sonnet-4.5 Gemma4-31B Gemma4-26BA4B Qwen3.5-35B-A3B Qwen3.6-35B-A3B STEM and Puzzle MMMU 82.3 79.6 80.4 78.4 81.4 81.7 MMMU-Pro 75.0 68.4 76.9* 73.8* 75.1 75.3 Mathvista(mini) 87.8 79.8 79.3 79.4 86.2 86.4 ZEROBench_sub 36.2 26.3 26.0 26.3 34.1 34.4 General VQA RealWorldQA 83.7 70.3 72.3 72.2 84.1 85.3 MMBench EN-DEV-v1.1 92.6 88.3 90.9 89.0 91.5 92.8 SimpleVQA 56.0 57.6 52.9 52.2 58.3 58.9 HallusionBench 70.0 59.9 67.4 66.1 67.9 69.8 Text Recognition and Document Understanding OmniDocBench1.5 88.9 85.8 80.1 74.4 89.3 89.9 CharXiv(RQ) 79.5 67.2 67.9 69.0 77.5 78.0 CC-OCR 81.0 68.1 75.7 74.5 80.7 81.9 AI2D_TEST 92.9 87.0 89.0 88.3 92.6 92.7 Spatial Intelligence RefCOCO(avg) 90.9 -- -- -- 89.2 92.0 ODInW13 41.1 -- -- -- 42.6 50.8 EmbSpatialBench 84.5 71.8 -- -- 83.1 84.3 RefSpatialBench 67.7 -- -- -- 63.5 64.3 Video Understanding VideoMME (w sub.) 87.0 81.1 -- -- 86.6 86.6 VideoMME (w/o sub.) 82.8 75.3 -- -- 82.5 82.5 VideoMMMU 82.3 77.6 81.6 76.0 80.4 83.7 MLVU 85.9 72.8 -- -- 85.6 86.2 MVBench 74.6 -- -- -- 74.8 74.6 LVBench 73.6 -- -- -- 71.4 71.4 * Empty cells (--) indicate scores not available or not applicable. Build with Qwen3.6-35B-A3B # Qwen3.6-35B-A3B is available as open weights on Hugging Face and ModelScope for self-hosting, and through the Alibaba Cloud Model Studio API as qwen3.6-flash . You can also try it instantly on Qwen Studio . The model can be seamlessly integrated with popular third-party coding assistants, including OpenClaw, Claude Code, and Qwen Code, to streamline development workflows and enable efficient, context-aware coding experiences. API Usage # This release supports the preserve_thinking feature: preserving thinking content from all preceding turns in messages, which is recommended for agentic tasks . Alibaba Cloud Model Studio # 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. Example code for chat completions API is provided below: """ Environment variables (per official docs): DASHSCOPE_API_KEY: Your API Key from https://modelstudio.console.alibabacloud.com DASHSCOPE_BASE_URL: (optional) Base URL for compatible-mode API. - Beijing: https://dashscope.aliyuncs.com/compatible-mode/v1 - Singapore: https://dashscope-intl.aliyuncs.com/compatible-mode/v1 - US (Virginia): https://dashscope-us.aliyuncs.com/compatible-mode/v1 DASHSCOPE_MODEL: (optional) Model name; 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. " "Set it via: export DASHSCOPE_API_KEY='your-api-key'" )

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-flash" , ) completion = client . chat . completions . create ( model = model , messages = messages , extra_body = { "enable_thinking" : True , # "preserve_thinking": True, }, stream = True )

reasoning_content = "" # Full reasoning trace answer_content = "" # Full response is_answering = False # Whether we have entered the answer phase print ( " \n " + "=" * 20 + "Reasoning" + "=" * 20 + " \n " )

for chunk in completion : if not chunk . choices : print ( " \n Usage:" ) print ( chunk . usage ) continue

delta = chunk . choices [ 0 ] . delta

# Collect reasoning content only 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

# Set your API key export DASHSCOPE_API_KEY =

# Start Qwen Code (interactive) qwen

# Then, in the session: /help /auth On first use, you’ll be prompted to sign in. You can run /auth anytime to switch authentication methods. Claude Code # Qwen APIs also support the Anthropic API protocol, meaning you can use it with tools like Claude Code for elevated coding experience: # Install Claude Code npm install -g @anthropic-ai/claude-code

# Configure environment export ANTHROPIC_MODEL = "qwen3.6-flash" export ANTHROPIC_SMALL_FAST_MODEL = "qwen3.6-flash" export ANTHROPIC_BASE_URL = https://dashscope-intl.aliyuncs.com/apps/anthropic export ANTHROPIC_AUTH_TOKEN =