OpenAI just launched the Decisions API, its answer to TypeSafe AI’s Jev. Both do the same basic thing…
George B. — Senior Software Engineer
OpenAI just launched the Decisions API, its answer to TypeSafe AI’s Jev.
Both do the same basic thing: instead of generating text, return a decision — predicate, choice or score — with probabilities.
Jev is cheaper: $0.042 vs $0.10 per 1M input tokens. OpenAI gives you images, much larger context, and claims up to 10x lower latency than using GPT-6 Luna through the Responses API.
Btw, Jev only launched on September 15. Three weeks later, TypeSafe says it was already processing ~1 trillion tokens/day and being used by ~25% of Fortune 500 companies.
That probably explains why OpenAI moved quickly. Databricks has already launched ai_decide too.
For spam detection, moderation, lead scoring, ticket routing or deciding which tool an agent should call next, paying an LLM to generate an answer may simply be unnecessary.
Jev looks hard to beat for high-volume text. OpenAI gets interesting when the decision involves images or huge amounts of context.
Now I want to see an independent benchmark of the two.
#AI #OpenAI #Jev #AIEngineering #LLM #AIAgents
5 reactions · 1 comments
Topics: #jev, #ai, #aiengineering, #llm, #openai, #aiagents
Comments
Schomer: Excellent structural breakdown, George. But the shift toward integrating these Decisions APIs (OpenAI or Jev) for automated tool-calling and ticket routing introduces a massive, systemic security blind spot [Other Google Activities, Other Claims]. The core problem is that whether an API returns generated text or a probabilistic score, it is still operating entirely as a software-layer application.