Anthropic widens lead on Vercel AI Gateway
Data from Vercel AI Gateway shows Anthropic's share of closed-weight model spend grew from 52% to 65% following the Opus 5.5 and Fable 5.2 releases. Meanwhile, open-weight models expanded dramatically across the platform, climbing from 2% to 29% of total gateway spend in 2026.
Anthropic's frontier model cadence has secured overwhelming revenue dominance among developers, while open-weight models are aggressively eating the commodity inference layer. The middle tier of closed-weight providers is getting squeezed from both ends as hybrid routing becomes standard practice.
- –Claude's dual release strategy—pairing high-reasoning Opus 5.5 with fast Fable 5.2—has made Anthropic the default closed-source stack, capturing nearly two-thirds of proprietary spend.
- –Open-weight models scaling from 2% to 29% of gateway spend in a single year proves that open models are no longer side experiments, but production cost-saving workhorses.
- –Squeezed between cheap open weights and Anthropic's frontier reasoning capabilities, rival closed-weight providers face severe margin pressure without standout differentiation.
- –Real-world gateway telemetry confirms developers are moving away from single-model setups in favor of routing commodity traffic to open weights while reserving Claude for complex reasoning.
DISCOVERED
7h ago
2026-11-19
PUBLISHED
7h ago
2026-11-19
RELEVANCE
AUTHOR
vercel