Step 5 Preview Enters OpenRouter’s Weekly Top Five

The completed-week ranking sets a weekly order

OpenRouter’s public rankings list text models by tokens processed through the OpenRouter platform. The page says its displayed date is the latest aggregated usage bucket, not the moment a reader loads the page. For October 5-11, that bucket offers a completed-week view of where work accumulated on this access layer. It does not measure all global AI use, and its token totals do not establish a model’s quality, revenue, or adoption outside OpenRouter. Within those limits, the ranking gives readers a clear, comparable weekly order.

DeepSeek V4.1 Flash ranked first with 41T tokens. GLM 5.3 Flash came second at 12.2T, followed by MiMo-V2.6-Flash with 11.5T. Step 5 Preview ranked fourth at 8.84T, and Hy4 preview ranked fifth at 8.28T. The ordering is the central result of this snapshot. DeepSeek V4.1 Flash had the largest volume in this particular bucket, while the next four entries formed a tightly bounded group between 8.28T and 12.2T tokens. The figures describe tokens processed by OpenRouter, not a verdict on the models themselves.

National Business Daily calculated 172T total OpenRouter tokens for October 5-11, up 3.61% week on week. The outlet calculated 86.74T for Chinese models, up 50.96%, and 18.97T for U.S. models, up 17.1%. Those country-group totals and growth rates are National Business Daily’s calculations from OpenRouter data. The ranking page supplies the model-by-model order; the newspaper’s calculations supply a broader tally for the same platform and week. Keeping those roles distinct makes the figures easier to read without stretching them into a measure of worldwide demand.

This new snapshot follows earlier EastFrontier coverage of OpenRouter traffic. The value of this update is its fresh order, especially Step 5 Preview’s place in the first five. A dated weekly bucket gives the reader a defined record rather than a claim that the order applies everywhere or will last indefinitely. It also makes the scope of the comparison plain.

Five Chinese models fill the ranking’s first five places

The five listed positions were occupied by Chinese models in this ranking: DeepSeek V4.1 Flash, GLM 5.3 Flash, MiMo-V2.6-Flash, Step 5 Preview, and Hy4 preview. That grouping is a statement about the weekly OpenRouter table. It should stay attached to the platform’s processed-token measure and its weekly scope. The list shows which named text models received the most routed tokens there during the completed period.

For readers comparing the rows, the first-place figure stands apart in size, while the other four entries make the first five a broader collection rather than a one-model result. Yet the numbers remain rank-table data. They do not show why a user selected a model, what task the token served, or how the model performed. No conclusion about global market share follows from a single platform’s weekly bucket. The practical use is narrower and still meaningful: it records the options that attracted the greatest token volume through OpenRouter in one finished week.

The result also offers a useful reference point for people navigating related EastFrontier coverage. This article keeps its attention on the present ranking and the calculations tied to it. OpenRouter’s list is organized around text models and tokens processed on the platform. Reading the table as a weekly traffic record preserves its usefulness while respecting the boundary of the data.

National Business Daily’s origin-group calculations add context without changing that boundary. Its Chinese-model total of 86.74T and U.S.-model total of 18.97T belong to the outlet’s calculation, as do the 50.96% and 17.1% week-on-week changes. The individual table rows answer a different question: which models appeared at the top of OpenRouter’s ranking. Together, the two views describe activity on OpenRouter only.

That distinction matters when the two data views are read together. The weekly rank uses individual text-model rows, ordered by the token volume processed through OpenRouter. National Business Daily’s separate calculation groups OpenRouter activity by model origin and compares it with the prior week. Neither view supplies an all-market measure. The first describes a ranking on one platform; the second describes the outlet’s calculation from that platform’s data. Keeping the two related but separate lets readers see both the named top-five order and the reported weekly totals without assigning either a claim it cannot support.

Step 5 Preview makes the next weekly bucket worth watching

Step 5 Preview’s fourth-place position is the change that gives this weekly record its immediate point of interest. At 8.84T tokens, it stood above Hy4 preview’s 8.28T and below MiMo-V2.6-Flash’s 11.5T. Its entry into the weekly top five distinguishes this account from earlier EastFrontier reporting on OpenRouter traffic. The present article need not assume a longer pattern to make that shift useful. It records the order published for a defined week.

The next completed OpenRouter week provides a simple follow-up. Readers can see whether this five-model arrangement appears again and whether Step 5 Preview remains among the most-used models on the platform. That comparison will be most useful if it keeps the same discipline as this one: read the page’s date as an aggregated usage bucket, compare tokens processed through OpenRouter, and avoid treating the result as a universal scoreboard.

For now, the weekly table gives a concise account of traffic on one visible access layer. DeepSeek V4.1 Flash held first place at 41T tokens. GLM 5.3 Flash, MiMo-V2.6-Flash, Step 5 Preview, and Hy4 preview followed in the stated weekly order, with Chinese models occupying all five positions. National Business Daily’s OpenRouter-based calculations add the week’s total and country-group context. The next bucket can show whether Step 5 Preview remains in the group or whether a fresh weekly order emerges.