← All models

Qwen3.7-Flash

Published Publisher Alibabaqwen3.7-flash-2026-07-15

Overview

Qwen3.7-Flash is a native vision-language model in the Qwen3.7 Flash series, served through Alibaba Cloud Model Studio. Alibaba describes improvements over Qwen3.6-Flash in multimodal understanding, object recognition, spatial perception, agent execution and multimodal coding. Suitable uses include visual understanding, text generation, reasoning, search agents and coding workflows. (alibabacloud.com)

Capabilities and limits

  • Accepts text, images and video; produces text. Supports function calling, structured outputs, prefix completion and context caching.
  • The context window is 1,000,000 tokens. Maximum input is 991,808 tokens, or 983,616 in thinking mode; maximum output is 131,072 tokens.
  • The snapshot does not support batch inference or fine-tuning. Web search support depends on region: it is supported in Beijing and Singapore but not in the documented Global deployments in Frankfurt, Virginia, Tokyo and Hong Kong. (docs.bailian.console.aliyun.com)

Release

The official release listing is dated July 21, 2026. The API ID identifies a snapshot dated July 15, 2026; the snapshot date is not the release-listing date. (alibabacloud.com)

Pricing

Original API prices below are for this specific snapshot, per 1 million tokens, checked November 16, 2026. The tier is determined by input tokens per request. Output charges include the answer and, in thinking mode, reasoning tokens. Promotions are excluded. Domestic and international offerings are separate and are not currency conversions. (docs.bailian.console.aliyun.com)

Input tokens China site: Beijing input / output (CNY) International site: Singapore, International scope input / output (USD)
Up to 32K 0.20 / 0.80 0.03 / 0.13
Over 32K to 256K 0.60 / 2.40 0.10 / 0.40
Over 256K to 1M 1.20 / 4.80 0.20 / 0.80

Cached input has separate charges. For the same three tiers, implicit cache hits cost CNY 0.04 / 0.12 / 0.24 in Beijing and USD 0.006 / 0.02 / 0.04 in Singapore, per million input tokens. Explicit cache creation and reads have separate rates in the official pricing pages. The Beijing snapshot has no free-token allowance listed. Do not apply batch pricing shown for the unversioned alias to this snapshot. (docs.bailian.console.aliyun.com)

Official sources