StepFun's Step 5 Preview, a mixture-of-experts (MoE) model with a 1M-token context window, has appeared on OpenRouter, a platform that gives developers access to a range of AI models through a single interface.
The listing was shared on Hacker News, where the post drew 82 points and 23 comments. The source material provides only the headline, the OpenRouter listing URL and the discussion thread; it does not include benchmark results, pricing details, or an official technical description from StepFun.
A 1M-token context window is large enough to process very long documents, extended codebases or lengthy conversation histories in a single prompt. Mixture-of-experts architectures activate only a subset of their parameters for each input, which can reduce the computing cost of running a model of a given capacity. Availability on a multi-provider platform such as OpenRouter typically lets developers test a model through a standard API without setting up their own infrastructure.
Readers interested in the model's real-world performance should consult the OpenRouter model page and the Hacker News thread, where users may share early impressions. Independent evaluations and StepFun's own documentation will be needed to assess claims about accuracy, speed and cost relative to comparable models.
Source: Hacker News · Summarized by HeadlinesBriefing