OpenAI Decisions API alternatives

Who this page is for

You are here because OpenAI Decisions is in limited preview and you need to ship classify, route, or agent next action today. Or you need calibrated confidence scores, multi-question batching, or Score/Noul types that OpenAI has not documented yet. Or you want an open-source path that runs locally.

This page covers the documented alternatives. Jev first (GA, most directly comparable), then where Decisions still fits, then other decision-shaped products where sources exist.

Primary alternative: TypeSafe Jev

TypeSafe Jev launched September 15, 2026 — two weeks before OpenAI's DevDay announcement — as a System One decision model. It handles the same classify, route, and agent next action jobs Decisions targets.

What Jev does that Decisions has not published

• Three question types: Choice (pick one), Score (where on a scale), Noul (probability 0–1). Mix them in one call.
• Per-option probabilities: every Choice answer lists a probability for each option, not just the pick.
• Confidence scores: your code can act on confident answers automatically and escalate the rest to a person or larger model.
• Multi-question batching: 60 questions cost about the same as one. Jev answers them all in parallel.
• Published API: POST /v1/systemone with Python and JavaScript SDKs. No preview access needed.
• Documented pricing: $0.042 per million input tokens, output free. Decisions-specific pricing: unpublished as of late September 2026.

Source: TypeSafe models

Builds shipping on Jev today

Duncan

@ephraimduncan

Built a model router with Jev by @typesafeai. Jev decides what model fits your request best and the request is sent to that model.

X·Routing and model choice

A model router on Jev

tamara

@tamarajtran

found the perfect use case for @typesafeai Jev: instant compaction in 2026, why is compaction still a summarization prompt? Jev can make it instant by scoring every tool call and dropping what’s irrelevant

X·Context and memory

Pick

Instant compaction for Claude

Gregor Zunic

@gregpr07

Breaking: Browser Use + Jev = Ultrafast ⚡ Findings flights took 7s and cost only $0.0039 🤯 > new action space every step > DOM state space > small LLM fallback to type (this video is at 1x speed btw) Built a tiny open source browser agent. try it below ↓

X·Search

Pick

Flight search with Browser Use

Run time

~7 s

Cost

~$0.004

Hassan

@nutlope

I used Jev to classify 1,018 AI research papers. The result: $0.08 total cost and 256ms median end-to-end latency per paper. The pipeline was: 1. Summarize each paper with DeepSeek V4 Flash 2. Send the title + summary + 24 possible topics to Jev 3. Use Jev to classify each paper 4. Visualize everything on http://1kpapers.com The summaries cost $3.99 on @togethercompute. The classifications cost $0.08 on @typesafeai. So for just over $4 of inference, I ended up with a pretty useful way to explore the top AI research papers from the past year. I think this is where things are heading: different models for different parts of the workflow, instead of using one model for everything. I’m running evals on the Jev classifications before replacing the current ones, but the site is already live: http://1kpapers.com

Live site·Documents and OCR

1kpapers

Documents

1,018 papers

Total cost

$0.08

Median latency

256 ms

More on the directory, grouped by job on Jev use cases.

When Decisions is still the better fit

Vision. OpenAI states Decisions accepts text or images. Jev takes text only (strings, JSON objects, or arrays of strings). If your workflow needs to classify screenshots, UI state, or computer-use frames without preprocessing, Decisions handles it once you have access.

Until then: preprocess images to text. OCR (Tesseract, cloud OCR APIs), vision-language models (Qwen2-VL, GPT-4o, Claude), or detectors (YOLO, segmentation models) feed text to Jev. The can Jev read images page lists routes and limits for each.

Other decision model peers

The category emerged in September 2026 with multiple vendors shipping decision APIs. Brief notes on documented peers; verify current state before choosing.

Cloudflare Clef

Cloudflare Clef is a System One decision model on Workers AI. Multimodal (text/JSON/images/video), Apache 2.0 open weights on Hugging Face, up to 64 questions per call, $0.24/M input per Workers AI docs.

Clef wins on vision and self-hosted open weights. Jev wins on documented per-token cost ($0.042/M vs $0.24/M) and proof builds. See Cloudflare Clef alternatives and Jev vs Cloudflare Clef.

Perplexity Decisions API

Perplexity Decisions on model pplx-decider-v1-27b. Handles noul/choice/score, 1–128 questions per call, images via base64, $0.04/M input (output free).

Perplexity wins on vision and slightly lower per-token cost ($0.04/M vs Jev $0.042/M — difference ~5%). Jev wins on proof builds. See Perplexity Decisions alternatives and Jev vs Perplexity Decisions.

OpenRouter Decisions API

OpenRouter branded a “Decisions API” endpoint that serves TypeSafe Jev and other decision models. This predates OpenAI's DevDay launch. Not the same product; specify which when discussing to avoid name collision.

Open-source path

For local or self-hosted alternatives, see open-source Jev alternatives. Examples: Laya at 421M parameters (86.5 decisions/second, ~9 ms P50 per author), Julia 1 at 144M (runs on CPU), and adapters that restrict next-token logits on models you already serve. No API call, no token bill, and the decision never leaves the device.

AWS Strands Decider

AWS Strands Decider was mentioned in the same category-validation window on X. As of early October 2026, publicly documented details are sparse. Check AWS announcements for current availability and schema.

How to decide

1. Do you need vision today? If yes and you have Decisions preview access, use it. If no access, preprocess to text and use Jev.
2. Do you need confidence scores or Score/Noul? Jev documents them; Decisions does not (as of late September reporting).
3. Do you need GA with published pricing? Jev is documented at $0.042/M input, output free. Decisions pricing: unpublished.
4. Do you need to run locally or self-host? See open-source alternatives.

For a detailed comparison: Jev vs OpenAI Decisions

Try it today

Next steps

• How to use Jev — first call, SDKs, gateways
• Jev pricing — cost per decision across real builds
• What is Jev — System One model, question types, limits
• Open-source Jev — local models that answer typed questions
• Agent prompts — wire Jev into Claude Code, Cursor, Codex

Common questions

What are alternatives to OpenAI Decisions API?

TypeSafe Jev is the primary generally available alternative with a documented API, probabilities, and confidence scores. For open-source paths, see decision models like Laya (421M parameters) and Julia 1 (144M). OpenRouter, Cloudflare, and other gateways may add decision-shaped endpoints.

Is TypeSafe Jev an OpenAI Decisions alternative?

Yes. Both handle classify, route, and agent next action jobs. Jev is generally available with a published contract (Choice, Score, Noul), per-option probabilities, and documented pricing ($0.042/M input, output free). Decisions is in limited preview as of late September 2026.

Why use Jev while Decisions is in limited preview?

Jev is GA with a published API, SDKs, and documented pricing. If you need calibrated confidence thresholds, multi-question batching (~60 questions cost about the same as one), or Score/Noul question types, Jev handles them today. No waitlist, no preview access needed.

When is OpenAI Decisions still the better choice?

When you need vision support. OpenAI states Decisions accepts text or images. Jev takes text only. For visual input (screenshots, UI state, computer-use frames), wait for Decisions access or preprocess images to text with OCR or a vision model first.

Is OpenRouter's Decisions API the same as OpenAI's?

No. OpenRouter branded a 'Decisions API' endpoint that can serve TypeSafe Jev and other decision models. It predates OpenAI's DevDay announcement. Specify 'OpenAI Decisions' vs 'OpenRouter Decisions' to avoid confusion.