China AI Native Industry Insights – 20260423 – Alibaba | Xiaomi | ByteDance | more
Alibaba unveils Qwen3.6-27B, Xiaomi debuts MiMo-V2.5-Pro, Volcengine’s Hermes launch. Discover more in Today’s China AI Native Industry Insights.
1. Alibaba releases Qwen3.6-27B open-source model with flagship-level coding capabilities
Alibaba released Qwen3.6-27B, a 27-billion parameter open-source dense model that outperforms its previous flagship Qwen3.5-397B-A17B across major coding benchmarks including SWE-bench Verified, Terminal-Bench, and SkillsBench. The model supports both thinking and non-thinking modes with a 262,144 token native context window extensible to over 1 million tokens. It is available under Apache 2.0 license on Hugging Face, ModelScope, and through Qwen Studio with API access.
Read more: https://qwen.ai/blog?id=qwen3.6-27b
Video Credit: The original article
2. Xiaomi releases MiMo-V2.5 Series models with MiMo-V2.5-Pro achieving frontier-level agentic capabilities
Xiaomi released the MiMo-V2.5 series including MiMo-V2.5-Pro and MiMo-V2.5 models, with the Pro version matching frontier models like Claude Opus 4.6 and GPT-5.4 on benchmarks including SWE-bench Pro 57.2 and Claw-Eval 63.8. MiMo-V2.5-Pro can autonomously complete complex professional tasks involving over 1,000 tool calls that would typically take human experts days or weeks. The standard MiMo-V2.5 model offers native omnimodal capabilities with image, video, and audio understanding, 1-million token context window, and pro-level agent performance at approximately half the cost of the Pro version. Both models are now available in public beta through Xiaomi’s API platform and will be open-sourced soon.
Read more: https://mimo.xiaomi.com/mimo-v2-5-pro
Video Credit: The original article
3. Volcengine Adds GLM-5.1 to Coding Plan and Launches Hermes Agent Plugin for ArkClaw
Volcengine has added GLM-5.1 to its Ark Coding Plan subscription, a multi-model coding package that also includes Doubao-Seed-2.0-pro, MiniMax-M2.7, and Kimi-K2.6. GLM-5.1 is described as capable of autonomous, continuous operation for over eight hours on long-horizon agentic tasks. Separately, Volcengine ArkClaw officially released a Hermes Agent plugin, allowing users to activate the open-source Hermes agent directly within ArkClaw without manual environment setup. The Hermes plugin leverages ArkClaw’s underlying OpenClaw architecture, enabling the agent to learn from conversations, generate reusable skills, and retain cross-session memory. Coding Plan Pro subscribers can access ArkClaw’s new Hermes capabilities directly through their existing subscription.
Read more: https://mp.weixin.qq.com/s/PGH22zYObkTHVZRXr9-I4Q
Video Credit: The original article
That’s all for today’s China AI Native Industry Insights. Join us at AI Native Foundation Membership Dashboard for the latest insights on AI Native, or follow our linkedin account at AI Native Foundation and our twitter account at AINativeF.