The state of open source AI
Which summary reads better? Pick one — models revealed after.Both summaries are AI-generated.
Open-source AI now powers over 50% of production tokens, with the top five highest-volume models on OpenRouter being open. This shifts cost control and customization to you—no per-token meters, full data sovereignty, and the ability to run offline or on-prem without vendor lock-in. The trade-off is operational overhead: open models lag in production tooling and trust, so expect to build or integrate more infrastructure to match closed-model deployment rates.
Open models now handle a majority of production tokens, with 79% of developers adding AI functionality using them, and the top 5 highest-volume models on OpenRouter being open; this shift enables enterprises to run AI models on their own hardware, reducing dependence on vendors and per-token costs, as seen with PwC fine-tuning an open model for the language of finance.