PRODUCTION USE CASE
A pricing service needs to explain a prorated subscription adjustment. Application code calculates the amount; the model receives the formula, verified intermediate values, currency rules, and customer context, then writes a concise explanation.
Production recipe
| Production choice | Recommendation | Why |
|---|---|---|
| API | POST /v1/chat/completions | OpenAI-compatible server-side request |
| Primary model | gpt-5.6-terra | Terra is the default for quantitative reasoning plus a clear customer-facing explanation. |
| Fallback model | gpt-5.4 | Use gpt-5.4 for formula explanation when all values are precomputed. Escalate genuinely difficult derivations to Sol, then still verify deterministically. |
| Escalation model | gpt-5.6-sol | Use only when the primary route fails the defined quality or complexity boundary |
| Output contract | Task-specific text or patch | A concise explanation whose amount, units, and rounding exactly match the independently computed result. |
A pricing service needs to explain a prorated subscription adjustment. Application code calculates the amount; the model receives the formula, verified intermediate values, currency rules, and customer context, then writes a concise explanation.
For open-ended quantitative analysis, the model can propose calculator or code-tool calls, but the final number is always recomputed and checked outside the model.
Send the request from your server. Replace the example content and placeholder tool schema with data and tools from your application.
curl https://app.llmfly.ai/v1/chat/completions \
-H "Authorization: Bearer $LLMFLY_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "gpt-5.6-terra",
"messages": [
{"role": "system", "content": "Explain the verified calculation using the supplied formula and values. Preserve currency, units, and rounding rules. Do not change any verified numeric input."},
{"role": "user", "content": "Explain this prorated subscription adjustment to the customer in three short steps."}
]
}'Terra is the default for quantitative reasoning plus a clear customer-facing explanation.
Use gpt-5.4 for formula explanation when all values are precomputed. Escalate genuinely difficult derivations to Sol, then still verify deterministically.
| Metric | Pass condition |
|---|---|
| Verified answer accuracy | The final answer matches an independently checked result on the target set |
| Unit and precision compliance | Units, rounding, significant figures, and uncertainty follow the task contract |
| Consistency across equivalent prompts | Equivalent formulations produce the same answer within allowed tolerance |
| Cost per verified solution | Model calls, tools, retries, and checks stay within the solution budget |
A concise explanation whose amount, units, and rounding exactly match the independently computed result.
Record the model ID, request ID, token usage, retries, validation result, and final disposition for every production run.
Evaluate the final answer and a concise, checkable explanation; do not depend on hidden reasoning.
Use deterministic tools when arithmetic, symbolic work, or simulation needs verification.
Compare cost per verified solution, including tools, retries, and validation.
Compare the primary and fallback models with the same requests, tools, and validation rules.