Choose Flash or Pro by the work you need to evaluate
Use the live catalog as the source of the exact model IDs that can be requested. For short, repeatable tasks, make Flash one candidate in your evaluation. For tasks where the quality of a longer or more involved answer matters, compare Pro against the same representative inputs.
This is a workload choice, not a capability guarantee. Measure the quality, latency, token mix, and failure cases that matter to your product before routing traffic.
Input and output are priced separately
A language-model call sends input tokens and receives output tokens. Providers normally publish a different price per one million tokens for each direction, so both numbers matter.
A long context with a short answer can be input-heavy. A short prompt that generates code or a long article can be output-heavy. Estimate both instead of multiplying only the visible answer length.
Use the live selling price
EidosStack publishes the current IDR selling price from an active price card. The card is locked before a request runs, so a price update cannot change a request already in flight.
Availability remains best-effort because upstream supply can change. Check the live catalog and use per-key budgets to bound spend.
Minimal request example
Use the exact model ID from the live catalog. The example only shows the request shape. Set the output limit and test the response with your own representative prompt before using it in production.
POST /chat/completions
Authorization: Bearer $EIDOSSTACK_API_KEY
Content-Type: application/json
{
"model": "deepseek-v4-pro",
"messages": [{ "role": "user", "content": "Ringkas teks ini." }],
"max_tokens": 500
}