AI Companies Develop Decision Models to Avoid Hallucinations

What happened
On September 15, TypeSafe AI introduced a new AI model called Jev, designed to make decisions quickly and avoid generating false information. Jev was released as a low-cost, high-speed alternative to existing large language models, and its launch prompted major technology companies and researchers to develop similar models. Independent tests confirmed Jev's fast response times and accuracy. The model has been widely adopted and imitated within weeks of its release.
Key details
- TypeSafe AI raised $870 million in funding at a $7.5 billion valuation.
- Jev delivers median response times of 150 milliseconds, compared to 1.89 seconds for OpenAI's GPT-6 Astra.
- TypeSafe AI charges $0.042 per million input tokens, with output tokens free.
- Cloudflare, Amazon, OpenAI, and Microsoft released their own decision models after Jev's launch.
- Decision models like Jev provide structured answers by choosing from predefined options.
Why it matters
Companies and developers now have access to much faster and cheaper AI models for decision-making tasks, reducing both operational costs and response times. Organizations that previously needed specialized teams and large datasets to build classifiers can now use off-the-shelf decision models for a wide range of applications. This change enables businesses and researchers to handle tasks such as content moderation and spam detection more efficiently, using models that are less likely to generate incorrect information.
