Structured Outputs alternatives for routing decisions

Who this page is for

You are here because you are using chat models with Structured Outputs or JSON mode for routing, classification, or agent next action, and the per-token cost or latency makes it impractical for high-frequency workflows. Or you need probabilities and confidence scores that JSON schema alone does not provide.

This page clarifies the category boundary. Structured Outputs and decision APIs solve different problems; neither replaces the other.

When Structured Outputs is the right tool

Occasional structured extracts. Parse a receipt, label a screenshot with complex annotations, extract fields from a document, convert prose to a schema your code expects. The output shape is complex or nested, and you run it occasionally (not thousands of times per minute).

Chat models excel here: they can generate arbitrary JSON that matches your schema, handle nested structures, and extract from long unstructured input. Structured Outputs guarantees the response conforms to your schema.

Cost is acceptable: if the workflow runs a few times per user or per hour, chat model pricing ($0.30–3.00/M input) is fine. The value is in the extraction, not in the per-call cost.

When a decision API wins

High-frequency classify/route/act. Model routing, agent next action, inbox triage, scoring feeds, tool selection. The question is which option, not extract this schema. You need probabilities and confidence, not just the pick.

Decision APIs excel here: they return per-option probabilities and confidence scores, batch 60–128 questions in one call at the cost of one, run at 70–500 ms latency, and cost $0.04–0.24/M input (7–75× cheaper than chat models per token).

Decision API options

• TypeSafe Jev: text-only, $0.042/M input (output free), GA, madewithjev proof builds. See how to use Jev.
• Perplexity Decisions: images via base64, $0.04/M input (output free), 1–128 questions. See Perplexity Decisions alternatives.
• Cloudflare Clef: multimodal (text/JSON/images/video), Apache 2.0 open weights, $0.24/M input. See Cloudflare Clef alternatives.
• OpenAI Decisions: limited preview on GPT-6 Luna, vision support. See OpenAI Decisions alternatives.

Key differences

Output shape

Structured Outputs: arbitrary nested JSON matching your schema. The model generates the structure.

Decision APIs: typed question (Choice/Score/Noul) with finite pre-defined options. The model picks from your list, does not generate new structure.

Probabilities and confidence

Structured Outputs: optionally returns token-level logprobs. No per-option probabilities for a typed decision. You must parse the JSON and interpret confidence yourself.

Decision APIs: return per-option probabilities and a confidence score for every question. Your code can act on confident answers automatically and escalate uncertain ones to a person or larger model.

Multi-question batching

Structured Outputs: one JSON response per call. To ask multiple questions, make multiple calls or nest them in one schema (more tokens, slower).

Decision APIs: 60–128 questions in one call cost about the same as one. Jev answers them all in parallel.

Latency

Structured Outputs: chat model generation latency, typically seconds for complex schemas.

Decision APIs: 70–500 ms per TypeSafe/Cloudflare docs. Purpose-built for classify/route/act loops.

Cost

Structured Outputs: chat model pricing, $0.30–3.00/M input depending on model (GPT-4o, Claude Sonnet, etc.).

Decision APIs: $0.04–0.042/M input (Jev/Perplexity), $0.24/M (Clef). 7–75× cheaper per token for classify/route workflows.

How to decide

1. Is the output shape complex or nested? Structured Outputs. Decision APIs only handle typed questions with finite options.
2. Is the workflow high-frequency (thousands/minute)? Decision APIs. Chat model cost and latency make high-frequency classify/route impractical.
3. Do you need per-option probabilities and confidence? Decision APIs return them natively. Structured Outputs: you parse and interpret yourself.
4. Do you need to batch many questions? Decision APIs: 60–128 questions in one call. Structured Outputs: one response per call.
5. Is cost-per-decision a constraint? Decision APIs are 7–75× cheaper per token than chat models for classify/route.

For detailed comparison: Jev vs Structured Outputs

Builds that show the job

Routing, classification, and agent next action — with the cost and latency each builder published on decision APIs.

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 Jev use cases. See also Jev vs an LLM for same-job both-ways comparisons.

Try it today

Next steps

• Jev vs Structured Outputs — side-by-side comparison
• How to use Jev — your first call, SDKs, gateways
• Jev vs an LLM — same job both ways, with published numbers
• TypeSafe Jev alternatives — Clef, Perplexity, OSS
• Agent prompts — wire Jev into Claude Code, Cursor, Codex

Common questions

What are alternatives to using Structured Outputs for routing decisions?

Decision APIs: TypeSafe Jev, Cloudflare Clef, Perplexity Decisions, OpenAI Decisions (preview). These return probabilities and confidence, batch multiple questions, run at 70–500 ms, cost $0.04–0.24/M input vs chat models at $0.30–3.00/M.

When should I use Structured Outputs vs a decision API?

Structured Outputs (chat with JSON schema): occasional structured extracts — parse receipts, label screenshots, extract fields. Decision APIs (Jev/Clef/Perplexity): high-frequency classify/route/act — agent next action, model routing, inbox triage, scoring feeds.

Can Jev replace Structured Outputs?

No. Different jobs. Jev answers typed questions (Choice/Score/Noul) for classify/route/act workflows. Structured Outputs extracts arbitrary schema from prose. Use Structured Outputs when the output shape is complex; use Jev when the question is which option.

What is the cost difference?

Decision APIs: $0.04–0.042/M input (Jev/Perplexity), $0.24/M (Clef). Chat models with Structured Outputs: $0.30–3.00/M input depending on model. For high-frequency classify/route, decision APIs are 7–75× cheaper per token.

Do decision APIs return probabilities like Structured Outputs?

Yes, but different. Decision APIs (Jev/Clef/Perplexity) return per-option probabilities and confidence scores for typed questions. Structured Outputs with logprobs can return token probabilities, but not per-option probabilities for a typed decision.