- The model topped OpenRouter’s Trending chart as Chinese AI models dominated the platform’s daily rankings
- With 600 billion parameters and competitive task-level costs, Step 5 Preview targets coding, financial analysis and other demanding workloads
Step 5 Preview, the latest flagship foundation model developed by Chinese AI model startup StepFun (阶跃星辰) topped OpenRouter’s Trending chart and ranked second worldwide on its daily leaderboard on October 10, drawing attention from developers in the global AI community.
Chinese models also filled the top five positions on the daily leaderboard, with DeepSeek V4.1 Flash in first place, followed by Step 5 Preview, Z.ai’s GLM 5.3 Flash, Xiaomi’s MiMo-V2.6-Flash and Tencent’s Hy4 Preview.
The rankings highlight the growing reach of Chinese models among developers using third-party platforms to access large language models.
OpenRouter aggregates models from different providers through a common interface. Its Trending chart tracks recent growth in usage, while the daily leaderboard ranks models by the total number of tokens they process.
A 600-billion-parameter model
StepFun plans to release Step 5 Preview’s full model weights on October 15. The model uses a sparse mixture-of-experts (MoE) architecture with 600 billion parameters in total, activating about 27 billion parameters per token.
It is designed for AI coding, software engineering, professional knowledge work and financial applications.
The model scored 44 on Artificial Analysis’s Intelligence Index, placing it among the top three open-source models globally, according to the company.
Balancing capability and cost
Its appeal to developers lies in the balance between capability and cost. AI researcher and DAIR.AI co-founder Elvis Saravia described the model’s performance as “impressive” after testing it, saying its capabilities were comparable to those of GLM-5.3 and Kimi K3.
StepFun says Step 5 Preview offers competitive pricing, placing it on the “Pareto frontier” of capability and cost — a balance that could make it worth integrating into commonly used coding agents.
According to the company, the cost of completing a task with the model is about one-eighth that of Anthropic’s Claude Opus 5.
Winning over developers
The model’s rise on OpenRouter offers a glimpse into how developers choose AI systems.
Complex tasks such as coding, financial analysis and deep research require more than strong benchmark scores: developers also need models that can handle demanding workloads without driving up operating costs.
For model developers, delivering more useful performance at a competitive price is becoming an increasingly important way to win adoption.
Step 5 Preview’s early traction suggests that Chinese AI companies are competing not only on technical capability, but also on the economics of putting their models to work.
Header image credit: StepFun’s official website
