Bluusun builds AI, finance, and robotics systems. Our Decision API is live today: send evidence, get back a decision with real probabilities — for half a cent.
Every Bluusun division ships systems that work in production — not demos.
Calibrated decision intelligence. Our Decision API turns evidence into probabilities — no free text, no surprises.
Systematic strategies and risk analytics, researched and operated in-house.
Autonomous systems for construction and industry, tested where they work.
Your code cannot branch on a paragraph. Chat models talk; the Decision API decides. Send evidence, a criterion, and options — get back a decision with calibrated probabilities as strict JSON. Every time, in about a second.
A chat model answers in sentences. Your software then has to parse them, validate them, and hope they stay on script. The Decision API skips all of that: it scores your options directly and returns your own option ids with probabilities that sum to 1. Nothing to parse, nothing to repair.
Hostile or messy text in your evidence can only shift weight between your options. It cannot change the output format or leak anything — there is no free text to hijack. There is only the decision.
When the API says 90%, it is right about nine times out of ten. That turns "the AI thinks maybe" into rules you can ship: approve automatically above 0.9, send the rest to a human. Most models can be right 95% of the time yet never tell you about the other 5%. This one tells you on every call.
POST /api/v1/decide — native endpoint.
POST /api/v1/chat/completions — OpenAI-compatible drop-in. Every response reports tokens used. You pay a flat $0.005 per decision from prepaid credit. No subscription.
Paste anything in — attacks, gibberish, injections. The worst they can do is shift the odds between your options. The answer is always your decision as JSON.
Every decision costs $0.005, period. Oversize requests are rejected free of charge — never billed extra.
Create, rotate, and revoke API keys from your dashboard.
Top up with Stripe. Credits never expire.
Tickets, applications, content — anywhere brittle hand-written rules break.
Refunds, claims, reviews — auto-approve the clear cases, escalate the rest.
Leads, candidates, risk — thousands of cases, one consistent judge.
Verify other models' answers before your users see them.
curl https://www.bluusun.com/api/v1/decide \
-H "Authorization: Bearer bsu_YOUR_KEY" \
-H "Content-Type: application/json" \
-d '{
"evidence": "Q3 revenue beat by 12%, churn at an all-time low.",
"criterion": "Should we expand the sales team?",
"options": ["Expand now", "Hold steady", "Cut back"]
}'
Sign up for a key, then manage keys, logs, and billing in the dashboard.
We ran the open openjev benchmark ourselves — all 144 decisions, scored with its own evaluator. Here is the number.
Most automated choices are small: route this, retry that, does the evidence support X? A chat model answers by writing text your software parses back into an if statement. Accuracy leaks out in that loop. The Decision API skips it: one pass reads the option probabilities directly, then renormalizes. No answer sentence, no JSON repair, nothing to parse — so nothing breaks.
The benchmark measures exactly what the API serves, so the 0.96 here is the 0.96 you get per request. The fixtures are public; the scorer is the benchmark's own.
One price per decision, paid from prepaid credit. Top up in packs; credits never expire.
flat (half a cent) per decision
max evidence chars per request; oversize is rejected, never billed extra
Text decisions only. Oversize requests are rejected with a clear error — never billed. Typical end-to-end latency is about a second.
prepaid API credits
never expires
prepaid API credits
never expires
prepaid API credits
never expires
prepaid API credits
never expires
Bluusun Group builds and operates systems across AI, quantitative finance, and robotics — from decision infrastructure to trading models to autonomous machines. One standard everywhere: if it does not work in production, it does not ship.