What does access to model weights mean for an enterprise that needs dependable AI? Sarbjeet Johal joins Dr. Efi Pylarinou on Fintech Conversations & Insights to discuss the practical economics and operational risks of open-weight models.
Ownership and operational continuity
The conversation distinguishes open-weight models from open source and examines the infrastructure needed to run them. Johal explores why possessing model weights does not, by itself, resolve questions of continuity, backward compatibility, or long-term business dependence.
Matching models to enterprise requirements
Johal and Pylarinou discuss model routing, experimentation, and the different requirements of startups and established enterprises. They also consider why systems of record and systems of engagement place different demands on AI accuracy and consistency.
The next enterprise AI architecture
The discussion looks ahead to specialized models, independent software vendors, and composite systems. It connects those possibilities with the continuing role of developers and the opportunity to make AI more broadly accessible.
Explore the conversation
- 00:45 — The real costs of open weights
- 05:28 — Open weights and open source
- 12:47 — Model routing
- 17:45 — Possession and continuity
- 19:26 — Backward compatibility risk
- 25:03 — Accuracy and systems of record
- 29:00 — Verticalized models and ISVs
- 30:18 — Composite systems
Video published on YouTube on September 2, 2026. Watch the original conversation on YouTube.