Daily AI CatchupOpen-Source AI's Token Share Surges to 62% at Vercel, Claiming Turf from OpenAI and Anthropic
Vercel's internal telemetry shows open-source AI model token consumption climbed from 28% to 62% in just two months, a dramatic swing away from proprietary providers like OpenAI and Anthropic. The shift reflects both cost pressure and improving open-source model quality making them viable for production workloads. Cheaper, locally-runnable models are cannibalizing API spending.
Why it matters
💻 Developer · Open-source models are now competitive enough to run in production for most tasks. You can build without vendor lock-in, tune models locally, and avoid surprise price hikes. The trade-off: support burden and fewer guarantees.
📦 Product · Your API spend is likely shifting to open-source whether you plan for it or not. This means either building open-source model ops in-house or doubling down on proprietary model differentiation.
🎨 Design · No UI change, but you'll increasingly see products built on open-source infrastructure. Expect faster iteration and more customization—but less polish on edge cases.
📈 Business · OpenAI and Anthropic's pricing power is weakening in the mid-tier. If you're not using their models for genuinely hard tasks, you're overpaying. This forces a race to the bottom or a pivot to specialized models.
🤔 Just Curious · Two months ago, open-source was a hobbyist option. Now it's half the market. The era of proprietary model dominance lasted about three years.
Sources: More data than open-source AI is taking share from OpenAI and Anthropic