Alibaba Releases Qwen3.8-27B: 27B Parameters Outperform Qwen3.7-Plus, Runs Locally
FlolightC · x · 2026-08-15
Alibaba has released the Qwen3.8-27B open-weight model, a native multimodal dense model with 27B parameters that outperforms Qwen3.7-Plus overall, excelling in real-world coding and office workflows. It supports 262K native context, extendable to 1M via YaRN, and is licensed under Apache 2.0. Running locally requires 17GB. The open weights for Qwen3.8-2.4T-A95B (Max-level) have also been released.
Related event: Alibaba Releases Open Source Qwen3.8-27B(40 posts)→
More from Models
- Qwen3.8-27B Local Test: Fact-Extraction F1 Ties Qwen3.6, Decode Speed Drops 16% — KitchenAmoeba4438 · 2026-08-15
- Qwen 3.8 27B 'overthinks' HTML/SVG tasks, generating 40k tokens — hurdurdur7 · 2026-08-15
- Qwen3.8-27B hits ~40 tok/s on dual RTX 3060, passes CUDA coding test — anderspitman · 2026-08-15
- Gemini 3.7 Flash leads Computer-Use Arena ahead of Claude Fable — cgarciae88 · 2026-08-15
- Information Abundance Paradox: long-context training can hurt short-task performance — DanielKhashabi · 2026-08-15
- Paid Gemini API returns 503 for 3.7 Flash constantly, while 3.1 Pro works fine — silverwoodz · 2026-08-15