← Sticker Wall MODEL · DETAIL

StepFun Terminal Scenario Model Licensing (Cockpit & Mobile)

1) Model licensing fees and customized deployment fees charged to terminal manufacturers; 2) Technical service fees base

MODEL

Key Fields

FIELD STAMPS

IndustryAI / LLM

RegionChina

ScaleMid-size

ChannelOnline

📌 Background

In 2026, the large model industry shifted from a parameter race to commercial implementation, with on-device and scenario-based applications becoming the key breakthrough areas for leading vendors. StepFun, which develops its own Step series of multimodal models, is generating scenario-based revenue through partnerships in smart cockpits and mobile systems, in addition to its open platform API. Following a Series B+ funding round of over 5 billion RMB during a financing winter, its commercialization path has drawn significant market attention.

👤 Target Customers

Terminal hardware manufacturers such as automotive OEMs and smartphone makers, as well as industry clients requiring embedded AI capabilities.

💰 Revenue Streams

1) Model licensing fees and customized deployment fees charged to terminal manufacturers; 2) Technical service fees based on device count or API call volume; 3) Revenue sharing from deep collaborations with automotive OEMs in in-vehicle interaction scenarios.

🧮 Cost Structure

Computing power investment for trillion-parameter model training, inference infrastructure, R&D investment for automotive-grade adaptation and terminal debugging, and sales and business development costs.

🛡️ Moat

Cross-modal interaction advantages derived from native multimodal end-to-end training, a high-low product mix of trillion-parameter MoE flagship models and low-cost Flash models, and first-mover positioning in partnerships with automotive and smartphone manufacturers.

🔑 Keys to Success

Securing benchmark orders from leading automotive and smartphone manufacturers to establish replicable solutions.
Continuously reducing lightweight model inference costs to match on-device computing constraints.
Building client stickiness through high-quality delivery and rapid responsiveness.

⚠️ Risks

Loss of licensing revenue if terminal manufacturers shift to in-house or open-source solutions.
IPO progress falling short of expectations, impacting capital availability and expansion pace.

🏢 Cases

Partnering with automotive OEMs to embed Step models into smart cockpit interaction systems.
Providing on-device model capabilities to mobile operating systems via an open platform.

📊 SWOT Analysis

Strengths

Native multimodal architecture provides superior user experience in cockpit voice-vision fusion scenarios.
Lightweight models like Step 3.5 Flash offer low inference costs, making them ideal for on-device deployment.

Weaknesses

Less than three years since inception, with a weaker track record in brand recognition and large-scale client delivery compared to industry giants.
High degree of customization in terminal partnerships leads to lower gross margins compared to pure API calls.

Opportunities

Rapid increase in penetration rates for smart cockpits and AI-powered smartphones in 2026.
Anticipated Hong Kong IPO provides capital support and brand endorsement.

Threats

In-house large model development by vendors like Huawei and Xiaomi squeezes the market space for third-party suppliers.
Open-source models like DeepSeek are driving down market expectations for licensing pricing.

Sources