Qwen3.7 Plus
Qwen3.7 Plus by Alibaba is a multimodal interactive hybrid agent combining GUI and CLI operation, vision input, preserve_thinking API, and a 1M-token context window.
Model Overview
Capabilities, design details, and architectural traits
Qwen3.7 Plus - Alibaba's Multimodal Interactive Hybrid Agent
Qwen3.7 Plus is officially described by Alibaba as a multimodal interactive hybrid agent - its defining identity. Built on the Qwen3.7 text backbone, it extends that foundation with full-modality vision input (images and video) while remaining a text-output-only model. Its core design goal is to unify GUI operation and CLI operation within a single agent loop, allowing it to read screens, interact with graphical interfaces, navigate mobile apps, and execute terminal tasks without switching between separate models or pipelines.
What Defines Qwen3.7 Plus as an Agent Platform
| Trait | Detail |
|---|---|
| Multimodal interactive hybrid agent | Officially named as such; operates across visual and text tasks in a single loop, not as a vision add-on to a chat model |
| Unified GUI and CLI operation | Perceives scenes, reads and clicks GUIs, navigates mobile apps, and runs terminal commands - all within the same agent session |
preserve_thinking API parameter | Retains <think> reasoning blocks across conversation turns, preventing the model from losing its reasoning chain between tool calls in long-horizon tasks |
| Cross-harness generalization | Documented to deliver stable performance across multiple agent frameworks without per-framework fine-tuning |
| Closed weights | API-only access via Alibaba Cloud Model Studio - a documented departure from the Qwen family's prior open-weight releases |
Context Architecture Designed for Long-Horizon Agent Loops
The model pairs a 1M-token context window with an internal 256K-token chain-of-thought budget. This allows a single agent session to hold a full codebase, screenshot history, and running plan simultaneously. The preserve_thinking parameter ensures the reasoning chain accumulated across turns is not discarded between tool calls - a documented architectural requirement for multi-step agentic workflows.
Benchmark Performance
Independent evaluations · Artificial Analysis
Accuracy & Capability Details
Independent evaluation data provided by Artificial Analysis. To view the latest benchmarks and full details, visit their official site.
Compare Models Side-by-Side
Evaluate specifications, pricing, and independent benchmark indices
| Model Details | AlibabaHost Qwen3.7 Plus | ||
|---|---|---|---|
| General Info | |||
| Provider | Alibaba | Anthropic | Anthropic |
| Release Date | June 1, 2026 | June 9, 2026 | May 28, 2026 |
| Knowledge Cutoff | — | — | — |
| Context & Limits | |||
| Context Window | 1M | 1M | 1M |
| Pricing (per 1M tokens) | |||
| Input Pricing | $0.40 Best Input Pricing | $10 | $5 |
| Output Pricing | $1.16 Best Output Pricing | $50 | $25 |
| Modalities | |||
| Inputs | textimage | textimagefile | textimagefile |
| Outputs | text | text | text |
| Benchmarks (0-100) | |||
| Intelligence Index | 39.0 | 59.9 Best Intelligence Index | 55.7 |
| Coding Index | 55.9 | 76.5 Best Coding Index | 74.3 |
| Agentic Index | 20.8 | 52.8 Best Agentic Index | 47.2 |
GPQA Benchmark
Graduate-level reasoning and expert Q&A evaluation.
Humanity's Last Exam
Extremely difficult logical reasoning and knowledge.
IFBench Instruction
Evaluation of strict instruction following.
Benchmark scores are independent evaluations sourced from Artificial Analysis. Intelligence, Coding, and Agentic indices reflect composite performance ratings (0–100).