Self-hosted GPU seconds — native MiniMax H3, LTX, and other open-weight runners — now use a 2× markup on Modal list price, down from 3×. Kling, Veo, Grok, and other partner APIs stay at vendor × 1.5. Runs already in flight settle at the new rate; unused reserve is refunded as before.
← All updates
Try it →
Self-hosted GPU seconds cost less
Open-weight clips (native H3, LTX) now bill Modal time at 2× list instead of 3×. Partner APIs are unchanged.