Jev cost calculator
Jev costs $0.042 per million input tokens, and output is free. A decision with 700 input tokens costs $0.0000294, so 100,000 of them a day cost $2.94 a day, or $88.20 over 30 days. Enter your own volume and request size to price Jev and 23 other hosted models side by side, cheapest first.
Cost for 100,000 decisions a day
Cheapest of the 30 selected: Celeris Decision at $8.40 a month, at 700 input tokens per call.
| Model | Per decision | Per 1,000 | Per day | Per month |
|---|---|---|---|---|
| Celeris Decision | $0.0000028 | $0.0028 | $0.28 | $8.40 |
| Cicero | $0.000007 | $0.007 | $0.70 | $21.00 |
| Span-02 | $0.000007 | $0.007 | $0.70 | $21.00 |
| Mercury Decide$0.02 per 1M input tokens, discounted from a regular $0.04 | $0.000014 | $0.014 | $1.40 | $42.00 |
| pplx-decider | $0.000014 | $0.014 | $1.40 | $42.00 |
| Kai | $0.0000147 | $0.0147 | $1.47 | $44.10 |
| Drex | $0.000021 | $0.021 | $2.10 | $63.00 |
| Drex 2.0 (Drex family) | $0.000021 | $0.021 | $2.10 | $63.00 |
| Drex Max Finance (Drex family) | $0.000021 | $0.021 | $2.10 | $63.00 |
| elephant-1 | $0.0000245 | $0.0245 | $2.45 | $73.50 |
| Clef-flash (Clef family) | $0.0000266 | $0.0266 | $2.66 | $79.80 |
| Decider 1 | $0.000028 | $0.028 | $2.80 | $84.00 |
| SubQ SnapEarly access | $0.000028 | $0.028 | $2.80 | $84.00 |
| Jev | $0.0000294 | $0.0294 | $2.94 | $88.20 |
| KevHosted on OpenRouter | $0.0000294 | $0.0294 | $2.94 | $88.20 |
| Microsoft-Decision-1Early access | $0.0000294 | $0.0294 | $2.94 | $88.20 |
| Tev1Early access | $0.0000294 | $0.0294 | $2.94 | $88.20 |
| pplx-decider-flash-v1.2 (pplx-decider family)Billed per question: up to 5 times the input | $0.000035 | $0.035 | $3.50 | $105.00 |
| pplx-decider-flash-v1.3 (pplx-decider family)Billed per question: up to 5 times the input | $0.000035 | $0.035 | $3.50 | $105.00 |
| Solar Decide Flash (Solar Decide family)Hosted on OpenRouter. $0.05 per 1M input tokens, discounted from a regular $0.10. Early access | $0.000035 | $0.035 | $3.50 | $105.00 |
| ClairicExtraction model. Early access | $0.00007 | $0.07 | $7.00 | $210.00 |
| d1Billed per question: up to 5 times the input | $0.00007 | $0.07 | $7.00 | $210.00 |
| Gemini Decision 1 | $0.00007 | $0.07 | $7.00 | $210.00 |
| mammoth-1 | $0.00007 | $0.07 | $7.00 | $210.00 |
| OpenAI Decisions API | $0.00007 | $0.07 | $7.00 | $210.00 |
| Solar DecideEarly access | $0.00007 | $0.07 | $7.00 | $210.00 |
| Clef-omni | $0.000105 | $0.105 | $10.50 | $315.00 |
| GLiDEBilled per question: up to 5 times the input | $0.00014 | $0.14 | $14.00 | $420.00 |
| d1-ultraBilled per question: up to 5 times the input | $0.000147 | $0.147 | $14.70 | $441.00 |
| Clef | $0.000168 | $0.168 | $16.80 | $504.00 |
Price per million tokens
Prices for every hosted model on this site, cheapest first, from each model's page. A model with a cheaper or different endpoint, such as Clef-flash, gets its own row. The model page has the source and the date each price was checked.
| Model | Maker | Input, per 1M tokens | Output | Status |
|---|---|---|---|---|
| Celeris Decision | Celeris | $0.004 | Free | Generally available |
| Cicero | Butler Compute | $0.01 | Free | Generally available |
| pplx-decider-flash-v1.2 (pplx-decider family)Bills each question as its own prompt | Perplexity | $0.01 | Free | Generally available |
| pplx-decider-flash-v1.3 (pplx-decider family)Bills each question as its own prompt | Perplexity | $0.01 | Free | Generally available |
| Span-02 | Respan | $0.01 | Free | Generally available |
| d1Bills each question as its own prompt | Liquid AI | $0.02 | Free | Generally available |
| Mercury DecideDiscounted from a regular $0.04 | Inception | $0.02 | Free | Generally available |
| pplx-decider | Perplexity | $0.02 | Free | Generally available |
| Kai | Hanzo AI | $0.021 | Free | Generally available |
| Drex | Nace.AI | $0.03 | Free | Generally available |
| Drex 2.0 (Drex family) | Nace.AI | $0.03 | Free | Generally available |
| Drex Max Finance (Drex family) | Nace.AI | $0.03 | Free | Generally available |
| elephant-1 | Baseten | $0.035 | Free | Generally available |
| Clef-flash (Clef family) | Cloudflare | $0.038 | Free | Generally available |
| Decider 1 | meraGPT | $0.04 | Free | Generally available |
| GLiDEBills each question as its own prompt | Fastino Labs | $0.04 | Free | Generally available |
| SubQ Snap | Subquadratic | $0.04 | Free | Early access |
| d1-ultraBills each question as its own prompt | Liquid AI | $0.042 | Free | Generally available |
| Jev | TypeSafe AI | $0.042 | Free | Generally available |
| KevOpen weights, hosted on OpenRouter | Jared Palmer | $0.042 | Free | Generally available |
| Microsoft-Decision-1 | Microsoft | $0.042 | Free | Early access |
| Tev1 | Together AI | $0.042 | Free | Early access |
| Solar Decide Flash (Solar Decide family)Served by OpenRouter. Discounted from a regular $0.10 | Upstage | $0.05 | Free | Early access |
| ClairicExtraction model | Cloudflare | $0.10 | Free | Early access |
| Gemini Decision 1 | $0.10 | Free | Generally available | |
| mammoth-1 | Baseten | $0.10 | Free | Generally available |
| OpenAI Decisions API | OpenAI | $0.10 | Free | Generally available |
| Solar Decide | Upstage | $0.10 | Free | Early access |
| Clef-omni | Cloudflare | $0.15 | Free | Generally available |
| Clef | Cloudflare | $0.24 | Free | Generally available |
How the maths works
Cost per decision is input tokens per call times the input price, divided by a million. A decision here is one call: the text you send plus the questions you ask about it. Per 1,000, per day and per month scale that figure, with a month counted as 30 days.
Output adds nothing. Every model in the table charges $0 for output. These models return typed values, not written text.
pplx-decider-flash-v1.2 (pplx-decider family), pplx-decider-flash-v1.3 (pplx-decider family), d1, GLiDE and d1-ultra bill each question as its own prompt. For those models the calculator multiplies the whole call by the number of questions. That is the most you would pay, since each question's own wording is only sent once.
A model page records its billing rule only when the maker's docs state one. Rows without one are priced as if the text you send is billed once per call. Mercury Decide and Solar Decide Flash (Solar Decide family) are priced at the discount their makers charge today, and the row gives the regular price.
The totals leave out free credit, free tiers, cached-input discounts and paid add-ons. Each model page lists the ones that apply. The Jev pricing guide works through these sums by hand.
FAQ
How much does one Jev decision cost?
Jev charges $0.042 per million input tokens and nothing for output. A decision with 700 input tokens costs $0.0000294, so a million of them cost $29.40. Only the tokens you send change the price.
Which decision model is cheapest per decision?
On price per token, Celeris Decision at $0.004 per million input tokens, so a million of the same 700-token decisions cost $2.80. Models can split the same text into different numbers of tokens, so check the token counts in your API responses before you compare bills.
Why does the number of questions per call matter?
For most models on this site, the published price bills the request once, however many questions it holds. pplx-decider-flash-v1.2 (pplx-decider family), pplx-decider-flash-v1.3 (pplx-decider family), d1, GLiDE and d1-ultra bill each question as its own prompt, so the text you send is paid for once per question. A model can also bill the text again when it splits a call's questions into groups, and its page says when. The calculator also flags a model when your question count is over its published limit per call.
Does the calculator include free tiers or free credit?
No. It uses the price each maker charges per token today. Where a maker marks that price as a discount, the row also gives the regular price. Some makers give new accounts free credit, offer a free tier or discount cached input, and each model page lists the ones that apply.
How do I compare a decision model with an LLM?
Open "Compare with an LLM", type the LLM’s input and output prices per million tokens and how many output tokens it writes per decision. That row uses the same input tokens as the other models and assumes one call per decision.
Every model is on the models page, and the ranking of decision models scores each model on independent benchmarks, price and access.