AI & Tech
95 Percent of Base10 Tokens Come From Custom Post-Trained AI Models
No Priors Podcast
Baseten CEO Tuhin Srivastava on the AI Inference Crunch, Custom Models, and Building the Inference Cloud
What was said
"95% of the tokens today are on the first business and almost all of them, there's probably a, yeah, for almost all of them, the customer's making some modifications to the model with their own data specialized for the use case."
The vast majority of AI inference workloads at Base10 involve custom models that companies have post-trained for their specific use cases, rather than vanilla open-source models. This demonstrates how the AI market is maturing beyond generic foundation models toward specialized applications with proprietary data and workflows.
Verification status
Attributed speaker claim
Egleze verified who said this and where. The underlying claim has not been independently established by Egleze.
Source and context
Egleze verified who said this and where. The underlying claim has not been independently established by Egleze.
- Programme
- No Priors Podcast
- Episode
- Baseten CEO Tuhin Srivastava on the AI Inference Crunch, Custom Models, and Building the Inference Cloud
- Clip time
- 13:47
- Original source
- Watch the source segment →
From this episode
No Priors Podcast