Alibaba's AI team Qwen has released new open model weights under the Apache 2.0 license with Qwen 3.8. The dense 27-billion-parameter model is designed to outperform the larger Qwen 3.7 Plus in coding and office tasks and natively processes up to 262,000 tokens of context. With this release, Qwen is targeting developers building local and agent-based applications.
SynthePulse Insight · AI deep reading
Version 1 · 1 source
Alibaba's Qwen team releases the Qwen 3.8 series of open-source models. The core 27B model claims to outperform the larger Qwen 3.7 Plus on coding and office tasks, with native 262K context expandable to one million.
On August 14, 2026, Alibaba's Qwen team released the open-source weights for the Qwen 3.8 series models under the Apache 2.0 license. The series includes the core model Qwen3.8-27B and the larger Qwen3.8-2.4T-A95B, positioned as the Max tier. All models are available on Hugging Face and ModelScope.
Qwen3.8-27B is a multimodal dense model with 27 billion parameters. According to the Qwen team, it outperforms the larger Qwen3.7-Plus on coding and office tasks. Additionally, the model has improved agentic capabilities, enabling more independent task planning and more reliable completion.
The model natively supports a context of up to 262,000 tokens, expandable to one million tokens via the YaRN method. Beyond text, it can process images and videos, including charts, documents, and multi-hour videos.
Qwen3.8-27B enables flexible thinking mode by default, but users can switch it per query. This feature provides flexibility for different tasks.
Qwen Cloud, Alibaba's AI service, will soon offer a hosted version with a one-million-token context for user convenience.
This report is based on secondary reporting from THE DECODER, which sourced information from the Qwen team's official statements on X. Performance claims (such as outperforming Qwen3.7-Plus) are source claims and have not been independently verified. Model specifications and availability information come from the same source.
The release of Qwen 3.8 marks a challenge by open-source models to larger models in performance. The Apache 2.0 license and long-context support may attract developers, but the performance advantages still require independent verification.
Primary report
Primary source