How do you choose between open and closed AI models?

Misha Laskin

Co-Founder, CEO · Reflection Ai

Control, Cost, Customization

Misha Laskin says ownership of intelligence over rental starts making sense at production scale for control of your data, lowering your costs, and customizing the model against your data.

MW

Matt White

Global CTO of AI · The Linux Foundation

Transparency

Matt White cites transparency, saying open models let you do your own safety testing and fine-tune them.

The answer starts at 7:54.

Vivek Vaidya

Co-Founder/CTO · Kana

Data Ownership and Usage Rights

Vivek Vaidya says the first thing to be clear about is who owns the data and who holds usage rights to it, since a B2B founder's data belongs to the customer, not the founder.

Eno Reyes

CTO & Co-Founder · Factory

Censorship, Capability, Longevity

Eno Reyes says to ask, for all models, what is potentially censored by the creators, whether the model can solve the problems you care about, and whether you can switch if it goes away.

Thomas Wolf

Co-founder - Chief Science Officer · Hugging Face

Need to Fine‑Tune on Own Data

Thomas Wolf says companies often need open-source models because they need to fine-tune on their own data, and open source is an easy way for a new company to build its own AI.

Matei Zaharia

CTO & Cofounder · Databricks

Flexibility and Control

Matei Zaharia says users want flexibility to move between models, which prevents lock-in, and that open-source models are strong if you want to control where they're served, hold on to data, or customize them.