Qwen3.7-Max Is Alibaba Betting the Model Is the Agent
Two days ago Qwen3.7 was just a preview label people spotted inside the Qwen Chat product, no weights, no API, nothing official. Now Alibaba has shipped Qwen3.7-Max as a full flagship, and the tagline tells you exactly where their head is at: The Agent Frontier.
Max is the reasoning, coding, and math workhorse of the line, while the broader 3.7 family is being positioned as a foundation for agents rather than a better chatbot. Look at the cadence and the story gets sharper. Qwen 3.5 in February, 3.6 in April, 3.7 in May. That is a major release roughly every six weeks, and it is happening under US export controls that were supposed to slow Chinese labs down, not speed them up.
Here is the read. Alibaba is doing what every frontier lab is quietly converging on right now, which is to stop selling you a model you chat with and start selling you the substrate your agents run on. Google said the same thing at I/O when it bragged that its new Flash beats last generation Pro on agentic benchmarks. Different companies, identical pivot.
The caveat is that Max is live in Qwen Chat but the full weights and API rollout are still the thing to watch, and a flagship is only as real as its availability. Worth tracking whether the open weights show up, because that is where Qwen has historically separated itself from the closed labs. https://qwen.ai/
← Back to all articles
Max is the reasoning, coding, and math workhorse of the line, while the broader 3.7 family is being positioned as a foundation for agents rather than a better chatbot. Look at the cadence and the story gets sharper. Qwen 3.5 in February, 3.6 in April, 3.7 in May. That is a major release roughly every six weeks, and it is happening under US export controls that were supposed to slow Chinese labs down, not speed them up.
Here is the read. Alibaba is doing what every frontier lab is quietly converging on right now, which is to stop selling you a model you chat with and start selling you the substrate your agents run on. Google said the same thing at I/O when it bragged that its new Flash beats last generation Pro on agentic benchmarks. Different companies, identical pivot.
The caveat is that Max is live in Qwen Chat but the full weights and API rollout are still the thing to watch, and a flagship is only as real as its availability. Worth tracking whether the open weights show up, because that is where Qwen has historically separated itself from the closed labs. https://qwen.ai/
Comments