Request and answer shape
Jev
state + choice / score / noul
Liquid d1
Same TypeSafe-compatible shape
Both can use the existing workbench input. Matching fields make integration comparable, not model behavior identical.
Decision model comparison · checked September 30, 2026
Same decision shape. Different gateway and price today.
Jev and Liquid d1 both answer typed decision questions. Their request and answer schema aligns, so the same Jev playground can run both.
The main practical difference is the serving route and listed price: this site currently calls Jev through OpenRouter at $0.042 per million input tokens, while Vercel currently lists Liquid d1 at $0. Compare both against the same labeled examples; no shared benchmark result for d1 is published here.
What the published details support
Compare the contract and bill separately from model quality
The request contract is compatible, but pricing comes from different gateways and a comparable JevBench result for Liquid d1 is not available in this comparison.
Upstream input price
Jev
$0.042 / 1M
Liquid d1
$0 / 1M currently listed
Jev price is the current OpenRouter listing used by this app. Liquid d1 is currently listed free on Vercel AI Gateway; check the live listing before budgeting.
Credits and 50% margin
Jev
600 / 1M · 30% gross margin
Liquid d1
1 minimum per successful call
At $0.0001 per credit, Jev needs 840 credits / 1M for a 50% gross margin. Liquid d1 has no positive cost base at a $0 listing, so it has a zero token rate plus the 1-credit call minimum.
When each path fits
Choose using the actual provider route, data, and cost boundary.
01
You want to trial a decision model with the current Vercel listing.
Try Liquid d1The Vercel AI Gateway public page currently lists $0 input and output prices. The app routes it through the Gateway TypeSafe endpoint with the configured server key.
02
Your production integration already uses Jev through OpenRouter.
Keep Jev while measuringThe Jev path in this app uses a published $0.042 per million input price. Compare latency, answer quality and confidence on a fixed dataset before migrating.
03
The probability controls a customer, payment, or safety action.
Run a labeled comparison firstA shared schema does not establish equal accuracy or calibration. Tune thresholds on held-out examples and preserve a review path for uncertain cases.
Serving and billing boundary
The app exposes one format while using separate provider connections.
Jev · OpenRouter connection
Use the existing model route
The app sends the shared request to OpenRouter and settles credits from actual input-token usage.
- Current provider rate: $0.042 per million input tokens; output is free.
- Current site rate: 600 credits per million input tokens, or $0.06 at the listed credit-pack rate.
Liquid d1 · Vercel AI Gateway
Use the same workbench over Gateway
The app forwards the TypeSafe-compatible request to Vercel AI Gateway using a server-side key.
- Current public listing: $0 per million input and output tokens; verify account pricing before relying on it.
- Current site billing: 0 token-based credits and a minimum 1 credit per successful request.
Request, provider and charge
The formats align; the service terms do not.
- Request format
- JevSystem One: state + choice / score / noul
- Liquid d1TypeSafe-compatible System One via Gateway
- Context listed
- Jev32K tokens
- Liquid d132K tokens
- Input / output price
- Jev$0.042 / $0 per 1M · OpenRouter
- Liquid d1$0 / $0 per 1M · Vercel public listing
- Current site credits
- Jev600 credits / 1M input tokens
- Liquid d11 credit minimum per successful call
Sources and scope
Prices are public gateway listings checked on September 30, 2026 and can change. Jev is compared at the OpenRouter rate used by this site; Liquid d1 is compared at its Vercel AI Gateway listing. No shared independent benchmark for d1 is asserted. The 50% margin example uses $0.0001 per credit and gross margin = (selling price − upstream cost) / selling price.