skip to content

Research • August 07, 2026

Chart of the Week: Open-Weight Models Win Tokens, Closed Ones Keep Cash

The walls may not be closing in on OpenAI and Anthropic as quickly as benchmarks and public sentiment suggest.

This article originally appeared in Galaxy Research's weekly newsletter. Subscribe to get timely insights delivered to your inbox every Friday morning.

Chinese open-weight AI models have turned inference into a commodity, calling into question the sustainability of the frontier labs’ 60%-80% markups. Yet the walls may not be closing in on OpenAI and Anthropic as quickly as benchmarks and public sentiment suggest.

Jesse Zhang, co-founder of Decagon, a maker of customer-service chatbots, recently argued that running open models requires technical finesse many corporations would rather pay to avoid. As the old adage goes, “no one gets fired for buying IBM.” Most enterprises could build the infrastructure in-house, but it's far easier to hand off the problem to Anthropic or OpenAI.

Data from OpenRouter, a single API endpoint that unifies inference providers from both open- and closed-weight models, supports this hypothesis. Since mid-May, open-weight models’ share of token usage has climbed nearly one-third to 75%. However, in terms of estimated dollars spent on the platform, closed-source models still crush open-weights.

Chart of the Week - inline 1

You are leaving Galaxy.com

You are leaving the Galaxy website and being directed to an external third-party website that we think might be of interest to you. Third-party websites are not under the control of Galaxy, and Galaxy is not responsible for the accuracy or completeness of the contents or the proper operation of any linked site. Please note the security and privacy policies on third-party websites differ from Galaxy policies, please read third-party privacy and security policies closely. If you do not wish to continue to the third-party site, click “Cancel”. The inclusion of any linked website does not imply Galaxy’s endorsement or adoption of the statements therein and is only provided for your convenience.