Choosing a Model
A short decision guide for picking the right delivery tier.
All tiers return the same grounded, source-consistent quality. Choose on latency, cost and plan — not on accuracy.
#Pick by what matters most
| If you need… | Use | Why |
|---|---|---|
| The lowest cost per call | Vedika Eco | Economy path from $0.01/query; ideal at high volume. |
| A balanced default | Vedika Standard | Deep synthesis at $0.01–$0.03; the right choice for most apps. |
| The newest Standard behaviour | Vedika Standard 2.5 | Set version:"2.5" on the standard path. |
| Sub-second, chat-grade latency | Vedika Swift | ~1,800 tok/s, 1–3s. Business+ only. |
| Maximum depth, latency no object | Vedika Pro Ultra | Deepest-synthesis profile of the standard path. |
#A quick heuristic
- Is this a live chat/typeahead experience? → Swift (needs Business+).
- Is cost the hard constraint at scale? → Eco.
- Is this a considered reading where quality leads? → Pro Ultra.
- Otherwise → Standard (optionally
version:"2.5").
Note
Latency figures are typical, not contractual. Per-query prices are USD; see Pricing and the live pricing page for INR and plan entitlements.
#Switching tiers is one field
Tier selection is per request, so you can route different workloads to different tiers with the same key — for example Swift for interactive chat and Eco for batch back-fills.
{ "question": "...", "birthDetails": { "...": "..." }, "speed": "fast" }