Nathan Lambert publishes a first data based analysis of the size of the open model economy.

Interconnects analyst Nathan Lambert has released a first research note on the open model economy. Fireworks disclosed serving more than 40 trillion tokens daily for a single segment alone, with total volume expected to pass 100 trillion tokens this fall.

Analyst Nathan Lambert, who runs the Interconnects newsletter, published a first research note on the open model economy, the commercial layer built around open weight AI models. He draws on figures from inference providers Baseten, Fireworks, and Together AI, plus the share of open models in OpenRouter token volume. At its first company conference, Fireworks disclosed serving more than 40 trillion tokens per day to customized models alone, a segment that has more than doubled since summer. Lambert expects the firm's total volume this fall to climb well past 100 trillion tokens.

Lambert benchmarks these figures against Google, the only major platform that discloses token volume in earnings reports. The growth curve for open models looks close to exponential, with an added jump after the GLM 5.2 model, which he credits with a step change in open agent systems. For companies weighing open versus proprietary models, this gives a rare quantitative anchor for how far the open ecosystem has scaled.

The data remains patchy. Hyperscalers such as Amazon Bedrock and Azure model garden likely carry a large share of open model inference but do not break out usage by model category. Lambert calls the analysis an early approximation and plans to continue the series as better numbers emerge. How large the gap to the proprietary market really is stays an open question.

This story was produced automatically from the source named above and checked by software before publication. The image is symbolic and shows neither the event nor a real person. How this paper is made