TypeSafe AI Launches Jev, A Non-LLM Model With Zero Tool-Call Errors
A co-developer of ChatGPT is now selling an architecture that abandons the left-to-right generation step those systems depend on, and TypeSafe AI is putting the claim on a price sheet rather than a benchmark chart.
Reporting from 1 source: GIGAZINE.
TypeSafe AI announced System One Models on September 15, 2026, along with Jev, the first model in the lineup. Jev skips sequential text generation and outputs a deliverable at once from a predefined structure. The company lists input at $0.042 per million tokens, output at $0, and says Jev matches GPT-5.6 Terra accuracy at about one-hundredth the cost.
TypeSafe AI is led by CEO Diogo Almeida, who was involved in ChatGPT's development. The company says Jev outputs results all at once according to a predefined structure instead of continuing text, and that no errors occur in its outputs. Every output carries a confidence score, and tool-call error rate is given as 0%. An early access program is open, with documentation posted for System One Models.
Synthesized by Yomimono from the 1 cited source below, including Japanese-language reporting where cited, then editorially reviewed before publishing.