Alibaba’s Qwen3.7-Plus cuts AI costs to $0.40 per 1M tokens with multimodal power

Alibaba’s Qwen team has unveiled Qwen3.7-Plus, the newest addition to its expanding family of large language models, offering multimodal capabilities at a significantly reduced price point. The model introduces support for text, video, and image inputs, enabling enterprises to process diverse data types without sacrificing performance. Priced at $0.40 per 1M input tokens and $1.60 per 1M output tokens, Qwen3.7-Plus undercuts many competitors while maintaining robust functionality.

A shift toward proprietary licensing

Unlike its predecessors in the Qwen lineup, which were released under open-source licenses, Qwen3.7-Plus is available exclusively through proprietary APIs and Alibaba’s Qwen Chat interface. This marks a notable departure from Alibaba’s prior strategy, which prioritized open-weight models trusted by major corporations like Airbnb. The move to closed licensing may disappoint enterprises that relied on the flexibility of open-source alternatives.

While the proprietary model limits customization, Alibaba positions Qwen3.7-Plus as a cost-effective solution for businesses requiring high-performance multimodal processing without the overhead of self-hosted deployments. The model’s pricing structure—60% lower than its immediate predecessor, Qwen3.7-Max—further reinforces its appeal to budget-conscious organizations.

Multimodal capabilities at a competitive price

Qwen3.7-Plus stands out for its ability to handle video analysis, image generation, and document processing, capabilities absent in text-only models like Qwen3.7-Max. This versatility allows businesses to automate workflows such as visual data extraction, real-time surveillance analysis, and cross-platform content creation.

Industry benchmarks suggest the model delivers competitive performance against rivals like MiniMax-M3, which recently launched at a limited-time discount, and DeepSeek’s v4-flash. Below is a comparison of leading AI models based on their input, output, and total token costs per 1M tokens:

Model               Input Cost  Output Cost  Total Cost  Provider
MiMo-V2.5 Flash        $0.10        $0.30        $0.40    Xiaomi MiMo
DeepSeek v4-flash      $0.14        $0.28        $0.42    DeepSeek
DeepSeek v4-pro        $0.435       $0.87        $1.305   DeepSeek
MiniMax-M3             $0.30        $1.20        $1.50    MiniMax
Qwen3.7-Plus           $0.40        $1.60        $2.00    Alibaba Cloud
Gemini 3.1 Flash-Lite  $0.25        $1.50        $1.75    Google

The model’s $2.00 per 1M token pricing places it among the most affordable options for enterprises seeking high-end multimodal AI without excessive costs.

Preserving reasoning in long-horizon tasks

For developers deploying autonomous agents, maintaining logical consistency across multi-step operations remains a persistent challenge. Qwen3.7-Plus addresses this with a 1-million token context window, allocating 256K tokens for internal chain-of-thought processing.

A key feature is the `preserve_thinking` parameter, which retains the model’s internal reasoning blocks across conversational turns. This capability, introduced in Qwen 3.6, ensures the model does not lose track of its analytical trajectory during extended tasks—such as code migration, automated testing, or data pipeline orchestration.

# Example: Enabling preserve_thinking in an API call
import requests

response = requests.post(
    "
    headers={"Authorization": "Bearer YOUR_API_KEY"},
    json={
        "model": "qwen3.7-plus",
        "messages": [{"role": "user", "content": "Analyze this codebase for vulnerabilities."}],
        "preserve_thinking": True  # Retains internal reasoning state
    }
)

This architectural improvement mirrors trends adopted by other AI labs, including Anthropic’s Extended Thinking and similar mechanisms in models like Claude Opus 4.8. By preventing state decay—where an agent’s focus drifts mid-task—Qwen3.7-Plus enhances reliability in enterprise automation scenarios.

The road ahead for Qwen’s ecosystem

Qwen3.7-Plus represents a strategic pivot for Alibaba, balancing cost efficiency with proprietary control. While the shift away from open-source may alienate some users, the model’s multimodal prowess and affordability position it as a compelling choice for businesses prioritizing out-of-the-box performance over customization.

As AI adoption accelerates, the demand for scalable, low-cost multimodal models will only grow. Whether Alibaba’s closed licensing model gains traction—or whether enterprises revert to open alternatives—remains an open question. For now, Qwen3.7-Plus offers a compelling middle ground between performance and pragmatism.