Summary
On July 30, 2026, Vercel updated GPT-5.6 access on its AI Gateway: GPT-5.6 Luna pricing dropped about 80%, Terra dropped about 20%, and Sol's fast mode got roughly 2.5x faster - all passed through with no token markup from Vercel.
What changed
Vercel adjusted GPT-5.6 pricing and performance on AI Gateway: Luna list pricing fell ~80%, Terra ~20%, and Sol fast mode throughput improved ~2.5x. Vercel states it applies no token-price markup, so the changes reach developers directly.
Why it matters
For teams routing model calls through AI Gateway, an 80% cut on a frontier tier materially changes the cost math for high-volume features and agent loops. Passing provider price and speed changes through without markup reinforces the gateway's pitch as neutral model plumbing rather than a margin layer.
Evidence excerpt
GPT-5.6 Luna pricing dropped 80%, Terra decreased 20%, and Sol's fast mode accelerated 2.5x - all without token price markups.