Qwen Releases Open Weights for Qwen3.8 Models
Qwen3.8-27B and a larger MoE model appear as open weights on Hugging Face.
Alibaba's Qwen account posted a countdown and performance details ahead of the Qwen3.8-27B release. Multiple observers noted the pre-release page and subsequent availability of the 27B dense multimodal model. Separate posts confirmed the open-weight drop of the 2.4T-parameter Qwen3.8-2.4T-A95B MoE model. NVIDIA AI Infrastructure publicly congratulated the team on the larger model. SGLang announced day-zero support with measured decode speeds on RTX 5090 and DGX hardware. Hugging Face links show the models listed for text generation under an Apache 2.0 license.
Qwen's biggest open-weight drop yet. 🚀 Meet Qwen3.8-Max: 2.4T params, 95B active, built to take a goal and come back with finished work. ⚡ 1M context. Adjustable reasoning. Parallel tool calls. 💻 Coded unattended for 10+ days, building a self-evolving harness from scratch…
Reactions
16 postsLA MI 
Mia@MiaAI_labQwen3.8-Max has been published on HF. A massive model with 2.4T params and 95B activated. Weights ~2.5 GB. You can download here are run on your 20 DGX Sparks.
- likes: 502
- replies: 48
- bookmarks: 85
- reposts: 37
YJ Yuchen Jin@Yuchenj_UWQwen3.8-Max weights just dropped on Hugging Face! 2.4T total parameters, with 95B activated. A good oss AI day.
- likes: 299
- replies: 10
- bookmarks: 26
- reposts: 21
NA NVIDIA AI Infrastructure@NVIDIAAIInfraMassive open-weight model drop! Congratulations to @Alibaba_Qwen on releasing the open weights for Qwen3.8-2.4T-A95B, a 2.4T parameter model with 95B active, designed for demanding reasoning and agentic workloads. Out of the box, the model achieves a throughput of 4K+…
- likes: 615
- replies: 30
- bookmarks: 59
- reposts: 53
ME merve@mervenoyannI have waited for open-sourced Qwen3.8 Max so bad for mostly this one sadly they released text decoder only but hey we can do a lot with that
- likes: 46
- replies: 4
- bookmarks: 11
- reposts: 1
RP Rohan Paul@rohanpaul_aiSo Qwen3.8-2.4T-A95B, the 2.4T-parameter open-weight mixture-of-experts model dropped on Huggingface.
- likes: 22
- replies: 6
- bookmarks: 1
- reposts: 4
