Build on the models behind Saanora.
One API, one key and one prepaid balance for every Saanora model. Clara, our decision model, is live today. Mark, the model people talk to in Saanora Chat, comes next.
Start with $0.20 of credit, no card needed · Clara: $0.10 per million input tokens, $0 for output
clara-1 · decision model
Send a record and the questions your software needs answered. Clara returns a probability for every option, in a fraction of a second. Probabilities, not paragraphs.
$0.10 per million input tokens · $0 output
mark-1 · conversational model · 272K context
The model people already talk to in Saanora Chat: writing, research, documents and code. Arriving on the same keys, credit and limits, priced per input and output token.
Pricing at launch
You ask three questions about a record. Clara answers all three.
- stateThe record. Any text or JSON, up to 16K tokens.
- failingYes or no.
- teamOne of the options you list.
- severityA point on your own scale.
166 input tokens · $0.000017 · 187 ms
A real response from the API, rounded to three decimals.
Small decisions, made thousands of times a day.
Routing, triage, risk, moderation, agent control. If the answer is yes or no, one of a list, or a point on a scale, Clara can answer it, for any record in any domain. A few things people ask:
- Is this a refund request?yes / no
- Which tool should the agent call next?choice
- How urgent is this ticket?score
- Does the invoice match the purchase order?yes / no
- Which queue owns this?choice
- Is there personal data in this message?yes / no
- Should a person approve this?yes / no
- Which policy does this break?choice
- How frustrated is the customer?score
- Is this transaction card testing?yes / no
- Is a required detail missing?yes / no
- Did the agent finish the task?yes / no
Level with Jev 1.13. Ahead of the OpenAI Decisions API.
On AgentDecisionBench's 3,939 agent decisions, Clara scores 91.9%, level with Jev 1.13 at 91.5%. On JevBench-mini it is within a point of the leaderboard's top model, inside the margin of error.
| Clara | 91.9% |
| Jev 1.13 | 91.5% |
| OpenAI Decisions API | 90.9% |
| Open-Jev-9B, leaderboard #1 | 83.9 |
| Clara, consistent mode | 83.0 |
Clara's scores are the live API, measured with each benchmark's official scoring. The other models' scores are each benchmark's published results, read in October 2026. Clara never trained on either benchmark. Full results
Describe the record once.
A ticket, a transaction, an agent's plan or a whole document, as text or JSON. No prompt to tune.
Ask what you need to know.
Up to 64 questions in one call, each yes or no, a choice from up to 255 options, or a point on a scale.
Branch on the numbers.
Every option gets a probability and they sum to one. Your thresholds live in your code, where you can test them.
per million input tokens
You pay for what Clara reads: the record and your questions. A request like the one above costs $0.000017.
- Output
- $0
- Per-decision fee
- None
- Smallest top-up
- $5
- Credit expires
- Never
- Failed requests
- Free
The boring parts, done properly.
Teams and roles
Invite people as owners, admins, developers or billing. Access is enforced in the database, not only in the interface.
Projects with spend caps
Separate keys and usage for each app or environment, and a monthly limit that stops spending at the line.
Every request on record
Request id, latency, tokens and cost for each call, by model, project and key. Export it as CSV. Failed calls are free.
Limits that grow on their own
Rate limits rise with your organization, from the first top-up to thousands of requests a minute. No forms.
Your records stay yours
We meter counts, tokens and cost. The records and questions you send are processed in memory and never written to our database.
Plain HTTP
JSON over HTTPS with request ids, typed errors and rate-limit headers. No SDK to install; curl works.
Most of the decisions inside software are small. Route this ticket. Flag that payment. Ask a person, or don't.
We kept watching teams send questions like these to chat models, then write code to dig the answer out of a paragraph. Clara skips the paragraph. It reads the record, answers exactly what you asked, and tells you how sure it is, so the rest of your system can decide what to do with that.
Mark, the model people talk to in Saanora Chat, comes to this API next, on the same keys and the same credit.
The Saanora team
Make your first decision today.
Sign up and you get $0.20 of credit, no card needed. Create a key, send a request, and add credit from $5 when you need more. The quickstart walks you through it.