Qwen3.7-Max

Qwen3.7-Max is the largest and most capable model in Alibaba's Qwen3.7 series. Alibaba positions it as a flagship model for agent-style work, with its main strengths in programming, office and productivity tasks, and long-running autonomous execution.

Versions and modalities

On Alibaba Cloud Model Studio the model ID qwen3.7-max currently points to the qwen3.7-max-2026-05-20 snapshot, a text-in, text-out model that runs in both thinking and non-thinking modes. A later snapshot, qwen3.7-max-2026-06-08, adds image and video understanding. An earlier preview, qwen3.7-max-preview (the 2026-05-17 snapshot), supports thinking mode only.

Capabilities and limits

Model Studio lists a 1,000,000-token context window, a maximum input of 991,808 tokens, a maximum output of 131,072 tokens and a chain-of-thought budget of up to 262,144 tokens. It supports function calling, structured outputs, prefix completion, context caching and, in most regions, built-in web search. Batch inference is offered in the China (Beijing) region. Fine-tuning is not supported on Model Studio.

Pricing

Pricing is usage-based per million tokens and varies by region. As of November 8, 2026, Model Studio lists $2.50 input and $7.50 output in Singapore, and $1.65 input and $4.951 output in US (Virginia), Germany (Frankfurt), Japan (Tokyo), Hong Kong and Beijing, with lower implicit-cache and explicit-cache-read rates. Rate limits also differ by region and snapshot.

Where to run it

Besides Alibaba Cloud Model Studio, the model is served by third-party platforms, including Together AI's serverless inference (model ID Qwen/Qwen3.7-Max, $2.50 input and $7.50 output per million tokens as of November 8, 2026) and OpenRouter, which routes requests to Alibaba endpoints. Hosts can differ in price, context length, quantization and rate limits, so the same model name does not guarantee identical behavior or cost.

Who it is for

Teams building coding agents, tool-using assistants and long-context workflows that want a large Qwen model through an OpenAI-compatible API. Alternatives include the newer Qwen3.8 Max, Kimi K3.1 Pro, GLM-5.3 and DeepSeek V4 models.

Pricing type: paid. Official model page: https://www.alibabacloud.com/help/en/model-studio/qwen3-7-max