What are the strengths of this model?

Supports diverse inputs including text, image, and video. Provides 1M context length for handling long sequences. Features mixed thinking and tool calling for enhanced functionality.

What are the weaknesses of this model?

Closed license with low transparency in development. Limited customization and modification options. Requires continuous support from the developer.

What are the best use cases?

Compound analysis combining images and text for insights. Automatic summarization of long documents or videos. Automated tasks via tool calling for efficiency.

AlibabaProprietary

Qwen3.7-Plus

Name: Qwen3.7-Plus
Author: Alibaba

Compare this model

Qwen3.7-Plus is a closed foundation model released by Alibaba in May 2026. It supports text, image, and video inputs, with 1M context length, 64K max output, mixed thinking, and tool calling capabilities. This multimodal intelligent agent can perform advanced reasoning and task execution.

Share:X Hatena LINE

Parameters

Undisclosed

Context Window

License

Proprietary

Release Date

2026-05-31

API Pricing

API pricing for this model is not yet available

Strengths

•Supports diverse inputs including text, image, and video.
•Provides 1M context length for handling long sequences.
•Features mixed thinking and tool calling for enhanced functionality.

Weaknesses

•Closed license with low transparency in development.
•Limited customization and modification options.
•Requires continuous support from the developer.

Use Cases

•Compound analysis combining images and text for insights.
•Automatic summarization of long documents or videos.
•Automated tasks via tool calling for efficiency.

Deep Analysis

Release

Late May 2026

Alibaba Qwen3.7 series; catalog date 2026-05-31

Context Window

1,000,000 tokens

Up to 131K token output

Pricing

$0.40 / $1.16 per 1M

Alibaba API; OpenRouter $0.32/$1.28; ~RMB 1.6/6.4 per M (0-256K)

Intelligence Index

53.3

Artificial Analysis; #40 of 534 models

GPQA Diamond

~88-90%

Epoch AI 87.9%; Dataconomy lists 90%

Strengths

・Strong reasoning at mid-tier price: GPQA Diamond ~88-90% and Tau2 93% put it near frontier reasoning models at a fraction of the cost.
・Full 1M-token context with 131K output, plus text, image, and video input - useful for long-document and multimodal workflows.
・Deeply integrated with Alibaba Cloud, DingTalk, and the Qwen open-source ecosystem; open variants are self-hostable.
・Coding Index 46.5 on Artificial Analysis makes it a credible everyday coding assistant.

Weaknesses

・Throughput is modest - about 53 tokens/sec and ~2.3s TTFT, near the bottom of the reasoning tier, so it lags on real-time interactive use.
・Verbose: it emitted ~110M output tokens on the Intelligence Index eval, inflating token cost versus tighter models.
・Proprietary weights for the Plus tier; only the open Qwen variants are self-hostable.
・Some reported benchmark figures (e.g. an HLE figure) are inconsistent across trackers; verify on your own tasks.

Competitor Comparison

Model	Arena	GPQA	Price
Qwen3.7 Plus (this)	-	~88-90%	$0.40/$1.16
DeepSeek V4 Flash	-	91.0%	$0.094 blended
GPT-5.6 Luna	-	91.6%	$0.450 blended
MiniMax M3	-	90.9%	$0.525 blended

Overview

Qwen3.7 Plus is Alibaba's balanced flagship of the Qwen3.7 family - a proprietary reasoning model with a 1M-token context, multimodal input, and near-frontier GPQA scores at $0.40/$1.16 per million tokens.

Benchmarks & Performance

Artificial Analysis gives it an Intelligence Index of 53.3 (#40 of 534) and Coding Index 46.5. Independently run GPQA Diamond lands at ~88-90%, Tau2 93%, IFBench 78%, and TerminalBench Hard 47%. Output speed is ~53 tokens/sec with ~2.3s TTFT - capable but not fast.

Detailed Comparison

It trades blows with DeepSeek V4 Flash and GPT-5.6 Luna on GPQA at a similar blended price, and undercuts Claude Opus-class models by 5-10x. The trade-off versus those peers is speed: Qwen3.7 Plus is deliberately a thoughtful, verbose reasoner rather than a low-latency one.

Community Feedback

Qwen users praise the 1M context and Alibaba Cloud integration for enterprise RAG; the main gripes are slow generation and token-heavy outputs. The open Qwen variants keep it popular for self-hosting.

Use Cases

Use it for complex business analysis, long-document Q&A, and multimodal understanding where you can tolerate a couple seconds of latency. For high-frequency chat or agent loops, pair it with a faster small model or a Flash-tier alternative to control cost and latency.

Latest News

Alibaba shipped Qwen3.7 Plus in late May 2026 (catalog date 2026-05-31) as part of the Qwen3.7 series alongside Qwen3.7 Max. Pricing settled at $0.40/$1.16 per million tokens on Alibaba's API, with promotional RMB rates through early July 2026. It remains a core offering in the Qwen Studio subscription.

Sources

Analysis generated: 2026-09-08