Rumor: Alibaba's Qwen to release a 125B-A6B MoE model, hailing the era of ~120B models
TheZachMueller · x · 2026-08-25
Developer Zach Mueller quotes a leak that @AlibabaQwen will soon release a 125B total / 6B active MoE model, remarking: "Welcome to the era of 120B models."
The spec continues Qwen's open-source MoE line, trading moderate total parameters for cost-efficient inference. This is an unverified third-party leak.
More from Models
- China's StepFun releases Step 5 Preview weights: 600B MoE, 27B active, 1M context — Dr_Singularity · 2026-10-15
- StepFun open-sources Step 5 Preview: 600B MoE with 1M context, No. 2 on OpenRouter — Dr_Singularity · 2026-10-15
- Google says teams across the company are using Gemini 4 Argon on its hardest engineering problems — CurieuxExplorer · 2026-10-15
- Perplexity's new embedding models: index with a 9B model, search with 0.6B — antoine_chaffin · 2026-10-15
- 797GB of n-gram counts in CPU RAM let 1B models match 57% more compute — PangWeiKoh · 2026-10-15
- Leak: Moonshot's Kimi flagship 'x4' done with pre-training, K3.1 in internal testing — lyraxana · 2026-10-15