Answer
Qwen model updates matter when a new branch changes open-model evaluation, coding performance, multimodal packaging, or English-accessible deployment paths. Verify the repo, model card, and docs before you test.
Key points
- Qwen is Alibaba's open-model initiative, hosted on GitHub and Hugging Face.
- Recent updates include the open-sourcing of Qwen3.6-35B-A3B (a 3B-active MoE model) and gray-release testing of HappyHorse 1.0 for video generation.
- No official Qwen4 or major architectural shift is documented in available evidence; activity emphasizes scenario-specific deployment over broad capability leaps.
What changed recently
- Qwen3.6-35B-A3B (MoE, 3B active parameters) was open-sourced on April 17, 2026.
- HappyHorse 1.0, a video generation model, entered gray-release testing on Qwen as of April 28, 2026.
Explanation
The evidence shows Alibaba prioritizing targeted, production-relevant releases: a sparse MoE model for efficiency and a video-generation model undergoing limited internal validation—not broad public release.
No release notes, versioned changelogs, or official QwenLM documentation updates are cited in the evidence; all observed activity is inferred from RadarAI briefs and aligns with infrastructure- and deployment-focused signals from April–May 2026.
Tools / Examples
- Qwen3.6-35B-A3B offers a trade-off: higher total parameter count (35B) with lower active compute (3B), potentially reducing inference cost per token.
- HappyHorse 1.0’s gray-release status means it is not yet publicly available or documented on QwenLM’s GitHub or Hugging Face repos.
Evidence timeline
OpenAI and Microsoft agree on multi-cloud decoupling to support IPO plans; Alibaba's HappyHorse 1.0 video generation model enters gray testing on Qwen; GitHub Copilot launches token-based AI credit billing.
Alibaba's HappyHorse 1.0 video generation model enters gray-release testing on Qwen; JD's JoyInside AI Hardware Innovation Contest sparks emotion-driven physical AI devices—signaling AI's shift from productivity tool to
Claude Opus 4.7 launches on Vercel AI Gateway; Alibaba open-sources Qwen3.6-35B-A3B (MoE, 3B active); JD.com and Mifeng unveil full-stack embodied AI data infrastructure.
Generative AI is rapidly shifting from a 'model capability race' to a contest over infrastructure sovereignty and deep, scenario-specific deployment: cost per token has become the core metric in NVIDIA's redefined techni
Embodied intelligence is accelerating toward engineering deployment; FluxVLA Engine emerges as the first open-source, standardized VLA foundation. MiniMax and Alibaba Cloud unveil the 'Model × Harness' paradigm and an en
Embodied intelligence is rapidly entering the 'deployment phase,' with Agibot proposing a new industry stage; dual breakthroughs in world models and VLA (Vision-Language-Action) engineering foundations—Alibaba's HappyOys
Sources
- RadarAI updates (evidence)
- QwenLM GitHub
- Qwen on Hugging Face
- RadarAI Methodology
- Sources & Coverage
- Signals Library
FAQ
Is Qwen4 released?
No evidence confirms a Qwen4 release. The latest referenced model is Qwen3.6-35B-A3B, open-sourced April 17, 2026.
Where can I find official Qwen release notes?
Official release notes are not linked in the evidence. Builders should monitor QwenLM’s GitHub and Hugging Face pages for authoritative updates.
Related
Last updated: 2026-05-09 · Policy: Editorial standards · Methodology