[ Compare ]
Jev vs OpenAI Decisions API: price, speed and accuracy compared
Published By JevHub
Jev and OpenAI's Decisions API do the same job: they answer a question about some input with a typed value and a probability instead of writing text. On published prices Jev is about 2.4 times cheaper on input; OpenAI's version also reads images and documents privacy controls. In the independent tests so far, accuracy is close and neither model wins everywhere.
Side by side
| Feature | Jev (TypeSafe AI) | OpenAI Decisions API |
|---|---|---|
| Status | Launched 15 September 2026 | Public beta since 6 October 2026; general availability expected "in the coming weeks" |
| Question types | Choice (up to 255 options), Score (2 to 10 levels), Noul (yes/no) | Choice, score, and predicate (the probability a condition is true) |
| What comes back | A typed answer with probabilities and a confidence value | A typed answer with probabilities and a confidence value; OpenAI does not explain how confidence is calculated |
| Input | Text only | Text and images (images as inline base64 data) |
| Input price | $0.042 per million tokens, output free (checked 22 September 2026) | $0.10 per million tokens, no output charge |
| Speed (each company's own claim) | 70 to 500 ms (TypeSafe) | "About 10x faster than the Responses API" (OpenAI); no absolute figure given |
| Size limits | 64,000 tokens per request, including all questions | Context window and rate limits are not stated in the documentation |
| Privacy controls | Not covered in the sources we checked | Zero Data Retention and HIPAA use for eligible customers; US and EEA/Switzerland regional processing |
| Models | Pin a version such as jev-1.13.0, or use the jev-latest alias | gpt-6-luna is the only model supported so far |
Prices and limits are as published by each company on the dates above and can change. Speed figures are claims, not measurements.
The short answer
Both products return typed decisions with probabilities rather than text, so they fit the same jobs: routing a support ticket, flagging a message, scoring a lead. Jev has been available since 15 September 2026; OpenAI's Decisions API entered public beta on 6 October 2026.
Three differences are documented. Price: Jev lists $0.042 per million input tokens and OpenAI lists $0.10, which makes Jev about 2.4 times cheaper per input token. That is our arithmetic from the published prices, not a measured cost for a whole task. Images: OpenAI's version accepts images; Jev takes text. Privacy controls: OpenAI documents Zero Data Retention, HIPAA use and regional processing; we did not find the same published for Jev.
OpenAI's own announcement
OpenAI introduced the Decisions API on 6 October 2026 with a short video. It describes the idea as turning text and images into actions in your app: you give the model an input, define a question and the possible answers, and get back a result your code can use.
What independent tests found
Few people have tested the two side by side so far, and each test uses a different task, so treat the results as early evidence. Here are the ones we could check.
Banking questions (StackToHeap, 8 October 2026). One developer ran 770 test questions from the public BANKING77 set through both models, with the pinned version jev-1.13.0 and gpt-6-luna. With only definitions of each category, Jev scored 81.17% and OpenAI 78.96%. With two worked examples per category, Jev scored 85.32% and OpenAI 84.16%. OpenAI answered faster on average (about 144 ms against 298 ms with definitions, and 194 ms against 520 ms with examples), and the estimated cost of the full run was lower for Jev ($0.069 against $0.106 with definitions, $0.183 against $0.302 with examples). The author's summary: Decisions for speed and larger inputs; Jev for cheaper, slightly more accurate classification. He notes the runs happened on different dates and under different load.
Survey answers (BTInsights, 8 October 2026). A market-research tool maker coded 2,617 real survey responses against human coders. Across all of them, at default settings, Jev scored 71.4 and OpenAI 72.3 (an F1 score out of 100); after tuning how strict each model is, Jev scored 74.0 and OpenAI 75.4. The authors report OpenAI cost about seven times as much, a much larger gap than the banking test found. The publisher sells a competing survey-coding product, so weigh this one accordingly.
A hands-on video (Mark Kashef, 6 October 2026). An AI consultant with about 86,000 subscribers walks through both APIs and runs three demos with the same question and the same evidence. His summary, in the video's description: if you want cheaper, Jev; if you want multimodal and faster, the Decisions API. This is one creator's opinion, not a controlled test.
What the tests do not tell you
Before you rely on any of this:
- Accuracy differences between the two were small and went both ways, depending on the task.
- Speed claims from both companies come with no published test conditions, and the independent timings were taken on different days.
- Neither company has published a matched head-to-head test.
- Your own questions and data will behave differently. Run both on a labelled sample of your own before you commit.
Which should you pick?
Based on what is documented today:
- Choose Jev if your input is text, you will run a lot of decisions and price per token matters, or you want a product that has been available since September.
- Choose OpenAI's Decisions API if you need to look at images, if you need the published privacy controls (Zero Data Retention, HIPAA, regional processing), or if you already build on OpenAI and want one vendor.
- Use both if you are unsure. Both accept the same style of question, so running the same labelled sample through each is a short experiment.
Keep reading
Where Jev fits, what it costs for your own volume, and how it differs from other decision models:
Common questions
- Is OpenAI's Decisions API better than Jev?
- Not clearly. In the independent tests so far, accuracy was close and each model led on some tasks. OpenAI's version adds image input and documented privacy controls; Jev is cheaper per token and has been available longer.
- Is Jev cheaper than the Decisions API?
- On published list prices, yes: $0.042 against $0.10 per million input tokens, about 2.4 times cheaper. In the one independent test that reported total cost on a labelled task, Jev's run cost less as well. Your real cost depends on how many tokens each of your requests uses.
- Can Jev read images?
- Jev takes text. OpenAI's Decisions API accepts text and images.
- Is the Decisions API available to everyone?
- It is in public beta, and OpenAI says it expects general availability in the coming weeks. Right now only the gpt-6-luna model supports it.
- Which one is faster?
- Both companies publish speed claims without test conditions. In the independent banking test, OpenAI's average response time was lower, but the two were measured on different days.
Sources
Written by JevHub from the sources below and checked against them on . Where a figure is a company's own claim, the article says so.
- OpenAI developer docs: Decisions (opens in a new tab)
- OpenAI on YouTube: Introducing the Decisions API (opens in a new tab)
- TypeSafe: Jev pricing and API reference (opens in a new tab)
- StackToHeap: OpenAI Decisions API vs Jev on BANKING77 (opens in a new tab)
- BTInsights: Jev vs OpenAI Decisions API on survey coding (opens in a new tab)
- Mark Kashef on YouTube: how Decisions compares to Jev (opens in a new tab)
More comparisons: Jev vs ChatGPT, Claude and other LLMs.