The Rise of Open Models in the Enterprise
summary
Calls out why enterprises move from API-driven OpenAI/Anthropic builds to self-hosted open models: quality wins on domain tasks (medical extraction, jargon transcription), latency and tail p99 issues, and exploding agentic call counts that multiply inference cost. Reports 100+ enterprise conversations and 40–50 production deployments by 2024, shows examples where fine-tuning and custom models beat closed APIs, and details the hard engineering: speculative decoding, prefix caching, Triton instability, replica cold start times, observability and four‑nines availability tradeoffs.
AI-written summary · about
chapters (6)
- 0:00Introduction to AI in enterprise
- 2:06Why enterprises adopt AI now
- 3:11Journey from closed to open models
- 6:46Cracks in closed model assumptions
- 10:43Challenges of open model adoption
- 15:42Build or buy infrastructure decision
related talks
10mThe Future for AI-First Enterprises is “Open” - How to Succeed and Not...
32mThe Rise of AI agents and open sources Ai models
11mOpen source GenAI vs. enterprise GenAI: Navigating the future of AI innovation
16mKeynote: Open Source for AI and AI for Open Source
31mStop Trusting a Black Box: The Economic Case for Open, Sovereign AI
14mReliable Agentic AI, the Open Way
27mRevolutionizing API Strategy in the AI Era
48mNavigating the Divide: Open vs. Proprietary AI