Alibaba Reveals Qwen4 Series: Max, Plus, Flash and 27B

Alibaba just put its next-generation model family on the main stage at Apsara Conference, and it is not a single release. The Qwen team laid out four models at once: Qwen4-Max, Qwen4-Plus, Qwen4-Flash, and Qwen4-27B.

The spread runs from cloud enterprise work down to self-hosted deployment, so there is a tier aimed at almost every way people already run Qwen today. Nothing has shipped yet, but the roadmap tells you where each one is headed.

What Happened

The four-line formation was announced on the Apsara Conference main stage:

• Qwen4-Max is the top tier, positioned against frontier models like Claude Opus. Alibaba is targeting hard reasoning, advanced math, and architecture-level coding with it.
• Qwen4-Flash goes after high-volume, low-latency workloads, a spot where it is set to compete with DeepSeek Flash.
• Qwen4-Plus and Qwen4-27B round out the lineup, covering the middle of the range and the smaller, self-hostable end.

Alibaba frames the whole thing as a push to cover every spectrum at once, from cloud enterprise deployments to models you run yourself.

Why It Matters

Most labs still announce one flagship and let the cheaper tiers follow later. Naming four at the same time means a Max user and a 27B user get told about their next upgrade on the same day.

The competitive part is the targeting. Qwen4-Max is being pitched directly at the reasoning and coding ceiling that Claude Opus holds, and Qwen4-Flash is aimed at DeepSeek Flash's efficiency lane rather than at a generic benchmark. If you pick models by cost per token or by whether you can host them yourself, those are the two lines worth watching.

Details on pricing, context window, and release dates are not part of this announcement. Until Alibaba publishes those, the lineup is a roadmap rather than something you can plan a migration around.