Skip to content
Trending storyDeveloping

OpenRouter reports open-weight models handled 71.7% of top-50 tokens in late September 2026

1 article1 sourcesince Oct 30Last article 2h ago ·

Overview

AISummary of 1 article

OpenRouter's blog defines open-weight models as models whose trained weights are published and available from many providers, not only from the model's maker.

In its data for the week of September 28, 2026, open-weight models handled 71.7% of tokens among the top 50 models on the platform. OpenRouter says these models cost about a quarter as much per token as closed-weight models. The post is a first-party account from OpenRouter, and the figures come from its own platform usage.

Written by AI from the articles below · updated Oct 30, 8:51 PM ET

Check the sources:

Article timeline

The articles in this story. Times are ET.

Oct 30
  1. OpenRouter BlogOfficialPick
    OpenRouter explains open-weight versus closed-weight models with usage and price data

    AIOpenRouter defines open-weight models as models whose trained weights are published, and says they are available from many providers rather than only the model's maker. Its data shows open-weight models handled 71.7% of top-50 model tokens in the week of September 28, 2026, and cost about a quarter as much per token.

Heat Trend

Not enough continuous observations to show a trend yet.