# Google and OpenAI advance decision models that return only scores

- Published: 2026-11-15T10:01:25.000Z
- Updated: 2026-11-16T22:02:56.211Z
- Section: Models
- Page: https://aidailyjournal.com/en/n/2026/11/15/xataka-google-and-openai-advance-decision-models-that-respond-with-numbers/
- Sources: [Xataka](https://xataka.com/robotica-e-ia/ia-rentable-2027-no-va-a-explicarnos-nada-solo-nos-dara-respuestas-como-0-82-0-16), [IEEE Spectrum](https://spectrum.ieee.org/typesafe-ai-hallucination-decision-models)

## Summary

Decision models receive context and a closed question—such as whether an email is urgent—and return an option with a score such as 0.82. Xataka reports that Google and OpenAI are among the companies moving in this direction, seeking faster and cheaper responses for decisions made by AI agents. The publication also warns that these systems could make the reasoning behind decisions such as prioritizing a complaint or blocking a payment less visible.

According to IEEE Spectrum, TypeSafe AI released Jev on 15 September and claims that the model is cheaper, faster and cannot hallucinate. The company raised US$870 million at a US$7.5 billion valuation, and the model inspired similar systems from Cloudflare, Amazon, OpenAI and Microsoft. Independent tests cited by the magazine found Jev had median latency of 150 milliseconds; TypeSafe AI charges US$0.042 per million input tokens, with no charge for output tokens.

## Why it matters (editorial interpretation)

Models that return only an option and a score could speed up narrow decisions by agents, but they may also make the reasoning behind actions such as prioritizing complaints or blocking payments less visible. This creates a trade-off between efficiency and the ability to assess those decisions.

_Summary written by AI Daily Journal in its own words, based on the sources above. For details, read the original articles._
