Alibaba unveils Qwen3.8-Omni-Flash multimodal AI model

Alibaba has introduced Qwen3.8-Omni-Flash, a new natively omni-modal AI model designed for professional productivity and AI agents. The model features a massive context window of up to one million tokens, allowing it to process diverse inputs including text, images, audio, and video within an integrated workflow.

Technical specifications include the ability to handle videos up to two hours long (or 2GB in size) and audio recordings up to three hours. The system supports speech recognition in 113 languages and dialects, with speech generation capabilities in 29 languages. In internal benchmarks, the model reportedly outperformed its predecessor, Qwen3.5-Omni-Plus, by more than 25% across 29 different evaluations.

To support adoption, the Qoder platform has announced a limited-time free access period for the Qwen3.8-Flash model, running from September 18 to September 30, 2026. During this window, users can utilize the model without incurring credit costs. Additionally, Qoder is offering daily login rewards of 100 credits to its users across various subscription tiers.