Open models hit 62% of token usage, and nobody saw it coming
Questions this post answers
What percentage of token traffic on Vercel's AI Gateway comes from open models now versus a few months ago?
Open models went from 28.4% of token traffic on Vercel's AI Gateway in June to 62% by August 22, according to figures shared by Vercel CEO Guillermo Rauch. He characterizes this as early-stage growth rather than a ceiling, since much of the developer tooling ecosystem (CLIs, IDEs, SDKs) is still hardcoded to specific closed models, a constraint that is expected to loosen over time. Track shifts like the open-model surge as you decide which models to build workflows around, on daily.dev.
How much money have companies saved by switching from closed to open AI models?
AT&T reportedly cut inference costs by 56% by routing to cheaper open models, and Coinbase saved around 50% doing the same. These enterprise cost reductions are cited as the primary driver behind the rapid shift in token usage toward open models, rather than any philosophical preference for openness. Compare real-world cost outcomes like these before committing budget to a model provider, via daily.dev.
How close are open-weight AI models to closed frontier models in quality?
Open models now trail closed frontier models by about 3.3% according to Stanford benchmark data cited in a SemiAnalysis chart, or by roughly four months according to Epoch AI's estimate. This is a sharp narrowing from a few years ago when the consensus was that open models could never catch up to closed labs. Keep tabs on benchmark gaps like this when picking a model for production, using daily.dev.