Alibaba releases Qwen 3.8 Omni Flash

Commenters debated Alibaba's Qwen 3.8 Omni Flash release, concentrating on cost, multimodal performance, and practical availability. ache highlighted a massive cost reduction versus Gemini using published pricing, while syntaxing claimed audio-visual performance approaches Gemini 3.8 Flash and that audio even exceeds it. conception praised Qwen as a “grounded” and reliable model, albeit slow and tied to Alibaba's access and stingy token plan, and tolugenius appreciated Qwen's broad range of smaller sizes for experimentation. cleaning and smallerfish compared it to other models like Astra and Opus 4.8, with smallerfish defending those as competent.

Opinion divides over release strategy, stability, and behavioral quirks. ache and others warned that open weight releases have slowed and doubted a full Qwen3.8-Omni-X or Qwen4 public release, while rubslopes questioned what “RL it to oblivion” would entail. spijdar described Flash-Next/Qwen4 reasoning as behaving weirdly and producing irrelevant internal thoughts, and nojs reported intermittent long latency and the need to resend requests. Some users celebrated the cost and capabilities; others expressed skepticism about long-term access, reliability, and the odd reasoning/response patterns observed.