Inworld opens consumer AI inference service built around interaction costs
The service offers hosted open models, dedicated deployments and cache-aware routing; Inworld says it already handles tens of trillions of tokens monthly.
By Ryan Merket · Published
Primary source: X / Inworld AI
Why it matters
Inworld is turning optimization work developed for interactive AI characters into an inference product for consumer apps. Its bet is that repeated conversational context and many model calls per user turn create enough cost and latency pressure to support a specialist serving provider.
Inworld AI opened its Inworld Inference service to developers on October 14th, 2026, selling infrastructure tuned for consumer apps where each slow reply and repeated prompt adds cost. CEO and co-founder Kylan Gibbs (@KylanGibbs), who previously led product work on large-language-model platforms at Google DeepMind, is taking the company beyond its original focus on AI characters and games and into the serving layer behind consumer AI products.…