Home
Alibaba Open Sources Qwen3.6-35B-A3B: 30B activated-parameter model delivers programming leap
On April 16, 2026, the Alibaba Qwen large model team released the sparse mixture-of-experts (MoE) model Qwen3.6-35B-A3B as open source, marking a key milestone in lightweight models for intelligent agent programming.

With 35 billion total parameters, the model activates only 3 billion during inference thanks to its sparse MoE architecture. In performance benchmarks, Qwen3.6-35B-A3B outpaces the dense 27-billion-parameter Qwen3.5-27B at a fraction of the computational cost, while also significantly outperforming its predecessor Qwen3.5-35B-A3B. It delivers logical reasoning and intelligent agent collaboration capabilities on par with large-scale models such as Gemma4-31B.

As a fully multimodal open-source model, Qwen3.6-35B-A3B excels in spatial intelligence and visual perception, achieving a RefCOCO score of 92.0. Some of its multimodal metrics already match the level of Claude Sonnet4.5. The model has been integrated into Qwen Studio and is available as the qwen3.6-flash API service on the Alibaba Cloud BaiLian platform. This API supports the preserve_thinking thinking chain retention function and seamlessly works with mainstream AI programming assistants like OpenClaw, Claude Code, and Qwen Code.
With rising demand for edge AI and automated intelligent agents, open-sourcing Qwen3.6-35B-A3B gives developers a high-performance, low-power option while signaling that MoE models with "small parameters, high intelligence" are becoming the new foundation for reshaping programming paradigms and multimodal interactions.
Related article
MiniMax 2.5 Tops Global Usage With Over $150M Monthly Revenue; M3 Flagship Version Set for H1 Release
Chinese large language models are generating intense momentum across the global developer ecosystem, driven by their reputation for cost-efficiency and engineering excellence.As of March 3, 2026, data from OpenRouter, the world’s leading API aggregat
Meta's Muse outpaces ChatGPT in early mobile launch
New data suggests Meta’s latest AI app, Muse, could become the social giant’s next major success. According to recent estimates from Appfigures, Muse has been downloaded on mobile devices more times in its first twelve days than ChatGPT did in the tw
TechCrunch Disrupt 2026: Clay’s Kareem Amin on the rise of the GTM engineer
Two years ago, GTM engineers were virtually nonexistent in most organizations. Today, the role is emerging across the tech sector, with professionals leveraging AI and automation to construct revenue engines that previously demanded multiple hires, d
Related Special Topic Recommendations
Comments (0)
0/500
On April 16, 2026, the Alibaba Qwen large model team released the sparse mixture-of-experts (MoE) model Qwen3.6-35B-A3B as open source, marking a key milestone in lightweight models for intelligent agent programming.

With 35 billion total parameters, the model activates only 3 billion during inference thanks to its sparse MoE architecture. In performance benchmarks, Qwen3.6-35B-A3B outpaces the dense 27-billion-parameter Qwen3.5-27B at a fraction of the computational cost, while also significantly outperforming its predecessor Qwen3.5-35B-A3B. It delivers logical reasoning and intelligent agent collaboration capabilities on par with large-scale models such as Gemma4-31B.

As a fully multimodal open-source model, Qwen3.6-35B-A3B excels in spatial intelligence and visual perception, achieving a RefCOCO score of 92.0. Some of its multimodal metrics already match the level of Claude Sonnet4.5. The model has been integrated into Qwen Studio and is available as the qwen3.6-flash API service on the Alibaba Cloud BaiLian platform. This API supports the preserve_thinking thinking chain retention function and seamlessly works with mainstream AI programming assistants like OpenClaw, Claude Code, and Qwen Code.
With rising demand for edge AI and automated intelligent agents, open-sourcing Qwen3.6-35B-A3B gives developers a high-performance, low-power option while signaling that MoE models with "small parameters, high intelligence" are becoming the new foundation for reshaping programming paradigms and multimodal interactions.
MiniMax 2.5 Tops Global Usage With Over $150M Monthly Revenue; M3 Flagship Version Set for H1 Release
Chinese large language models are generating intense momentum across the global developer ecosystem, driven by their reputation for cost-efficiency and engineering excellence.As of March 3, 2026, data from OpenRouter, the world’s leading API aggregat
Meta's Muse outpaces ChatGPT in early mobile launch
New data suggests Meta’s latest AI app, Muse, could become the social giant’s next major success. According to recent estimates from Appfigures, Muse has been downloaded on mobile devices more times in its first twelve days than ChatGPT did in the tw
TechCrunch Disrupt 2026: Clay’s Kareem Amin on the rise of the GTM engineer
Two years ago, GTM engineers were virtually nonexistent in most organizations. Today, the role is emerging across the tech sector, with professionals leveraging AI and automation to construct revenue engines that previously demanded multiple hires, d











