Jev alternatives: OpenAI, AWS and Cloudflare just launched decision models

Jev is a System One model: it answers a defined question about some input with a typed value and a probability, rather than writing text. Until this month, it was the only widely used model of that kind. Between 1 and 9 October 2026, three much larger companies each released their own.

OpenAI released a Decisions API in public beta on 6 October 2026. Requests go to /v1/decisions, and the API turns text or images into typed answers. gpt-6-luna is the only model that supports it so far. OpenAI says general availability is expected in the coming weeks.

OpenAI describes it as about 10 times faster than its Responses API. At the time of writing we could not find a published benchmark behind that comparison.

Strands Decider comes from Strands Labs, the experimental arm of AWS's open-source Strands Agents project. Coverage dates its release to 1 October 2026. The flagship model has 1.9 billion parameters, is built on Qwen3.5-2B, and is open source under the Apache 2.0 licence. You install it with pip install strands-decider and run it on your own hardware.

It answers the same three kinds of question as Jev (Choice, Noul and Score), each with a confidence value. On the public set of the JevBench benchmark, its own page reports 77.9% accuracy (180 of 231 tasks) and a median response time of 115 ms on a single RTX 3090 graphics card. These are the project's own measurements. It is an experimental lab project, not a managed AWS service.

Cloudflare introduced Clef and the smaller, faster Clef-flash on 1 October 2026. Both are open-weight under Apache 2.0 and also run on Cloudflare's Workers AI. Clef is built on a 27-billion-parameter base model and Clef-flash on a 9-billion one, each with Cloudflare's own training on top. Clef can also read images. Cloudflare says both are compatible with the Jev API.

In Cloudflare's own tests, the two models together lead on 7 of the 10 benchmarks in its table (Clef on 4, Clef-flash on 3), and Jev leads on 2. Cloudflare's measured median response time for Clef-flash is 38.8 ms against 524.1 ms for Jev, roughly 13 times faster, with Clef at 209 ms. Cloudflare also launched a service for fine-tuning decision models on a customer's own data.

Three caveats apply to all of the above:

• Speed and accuracy comparisons are mostly published by the companies releasing the models.
• Jev is a closed model; TypeSafe has not released its weights, while two of the three newcomers are open.
• Independent head-to-head tests on real workloads are still thin. We will update this article as they appear.

We compare Jev with each of the new models, on price, speed, features and what independent tests found:

• Jev vs OpenAI Decisions API
• Jev vs open decision models: Strands Decider and Cloudflare Clef

The category Jev opened now has serious competition, including options you can run yourself. If you are choosing for a new project, test your own question on more than one model; the use cases on JevHub show the kind of questions this class of model handles well.