Alibaba's Qwen team previewed Qwen3.8-Max-Preview on July 19, 2026 — timed to land during the World Artificial Intelligence Conference in Shanghai, and just two days after Moonshot AI's Kimi K3 open-weight launch. The timing isn't subtle: China's two biggest open-model labs put out flagship announcements within 72 hours of each other.
The headline numbers
Qwen3.8-Max-Preview weighs in at 2.4 trillion parameters and is Qwen's first multimodal model above the 1-trillion mark — natively handling text, images, video, and documents. Alibaba's own claim is that it's "second only to Fable 5" among the models it benchmarked, and that it beats its predecessor, Qwen3.7-Max, on coding, full-stack development, data analysis, and office workflows.
Preview, not release
Unlike Kimi K3 — which is heading toward open weights on July 27 — Qwen3.8-Max-Preview is, for now, a closed preview. It's being sold through Alibaba's Token Plan, Qoder, and QoderWork at 10% of standard pricing during the preview window, with steeper off-peak discounts for individual-plan users. Alibaba says an official version will follow and be open-sourced, at which point the developer community gets direct access to the weights. Until then, this is a taste of where the next Qwen flagship is heading rather than something you can self-host today.
Context: a very fast week
Qwen3.8-Max-Preview lands in the same stretch as Kimi K3's 2.8T-parameter open-weight launch — two Chinese labs pushing frontier-scale multimodal models into the public conversation within days of each other, both explicitly measuring themselves against the same closed-frontier bar (Fable 5, GPT-5.6 Sol). Whatever the benchmark numbers hold up to once wider testing happens, it's a clear signal of how fast the pace has gotten.
Links
- Pricing/Token Plan: QwenCloud Token Plan