Qwen3.7-Plus

// Overview

Qwen3.7-Plus is Alibaba's Qwen team's multimodal agent model in the Qwen-Plus line, announced on June 2, 2026 and generally available from June 1, 2026 after a short public preview. It bolts image and video understanding onto the text-only Qwen 3.7 backbone and is positioned as the lower-cost sibling of the flagship Qwen3.7-Max — Alibaba lists it at roughly one-sixth the per-token price of Max.

Unlike a text model with a vision adapter, Qwen3.7-Plus is built to operate as an interactive agent: it perceives real-world scenes, reads screens and grounds clicks on GUIs, writes code from visual references, navigates mobile apps end to end, and answers questions about video frames — blending GUI and command-line actions inside a single agent loop with tool calls, self-testing, and autonomous iteration. It is a perception model only: it reads images and video but returns text, not generated pictures.

The model carries a 1-million-token context window and is exposed as the API endpoint qwen3.7-plus on Alibaba Cloud Model Studio (DashScope), reachable through OpenAI-compatible chat-completions and responses APIs across Beijing, Singapore, and US-Virginia endpoints, and resold through aggregators such as OpenRouter. It is proprietary and API-only — no open weights have been published.

// Benchmarks

1. ScreenSpot Pro (GUI grounding)79%
2. AndroidWorld (mobile agent)81%
3. Terminal-Bench 2.0 (agentic terminal)70.3%
4. GPQA Diamond (STEM reasoning)90.3%
5. Artificial Analysis Intelligence Index39index

Scores on a 0–100 scale (25-point gridlines); higher is better. Each benchmark links to its published source.

// Pricing

Singapore (international) region, non-thinking mode, for prompts up to 256K tokens. Above 256K tokens the rate rises to $1.20 input / $3.60 output per 1M. Thinking-mode output is billed higher ($4 / 1M up to 256K). Pricing is tiered by request length.

Pricing source ↗

// Strengths

• GUI grounding and on-screen agent control: 79.0 on ScreenSpot Pro and 81.0 on AndroidWorld (vendor-reported)
• Very large 1M-token context window for long documents, codebases, and multi-turn agent traces
• Native image and video input alongside text — reads screens, frames, and document images
• Strong agentic tool use, self-testing, and autonomous iteration inherited from the Qwen 3.7 agent backbone
• Low per-token price relative to flagship agent models (about one-sixth the cost of Qwen3.7-Max)
• OpenAI-compatible API and multi-region availability make integration straightforward

// Best for

• Computer-use and mobile agents that read screens and click the right UI element
• Long-context document, codebase, and transcript analysis up to 1M tokens
• Visual question answering over screenshots, document images, and video frames
• Coding from visual references and UI mockups with built-in test-and-iterate loops
• Tool-calling agent workflows that mix GUI and command-line actions
• Cost-sensitive multimodal deployments that need vision without flagship-tier pricing

// How to access

// Qwen-Plus (multimodal agent) — every version

The full lineage of the Qwen-Plus (multimodal agent) line, newest first. Every version has its own page — click any to compare specs, benchmarks and pricing.