April 28, 2025
Alibaba Qwen3 released
Alibaba open-weighted Qwen3 on April 28, 2025 — hybrid thinking and non-thinking modes across dense and MoE models up to 235B parameters.
What it was for
Qwen3 matched or beat o1, DeepSeek R1, and GeminiGoogle's family of multimodal AI models — successors to PaLM and competitors to GPT and Claude. 2.5 on several public benchmarks while supporting 119 languages. Hybrid reasoningStep-by-step logical thinking in AI models — chain-of-thought before answering hard problems. let developers toggle extended thinking per task — making Alibaba a top-tier open-source alternative alongside DeepSeek and Meta.
Companies
- Alibaba
Why it's here
Qwen3 put Alibaba firmly in the open frontier model race with hybrid reasoningStep-by-step logical thinking in AI models — chain-of-thought before answering hard problems..
Why it mattered
Chinese open-weight models became credible defaults for global self-hosted deployments.
What it solved
TeamsMicrosoft Teams — chat, video calls, and Office integration for workplace collaboration. wanted o1-class reasoningStep-by-step logical thinking in AI models — chain-of-thought before answering hard problems. with open weights and flexible thinking budgets.
Media
- ImageAlibaba Group
Zeng Liansong, Public domain, via Wikimedia Commons
Related
- DeepSeek R1 releasedJanuary 20, 2025
- DeepSeek V3 releasedDecember 26, 2024
- OpenAI o1 reasoning models releasedSeptember 12, 2024