Models & Pricing
The live model catalog — every modality, capability badges, and how sats pricing works per model.
The AI Gateway catalog is synced automatically from upstream — new models appear and retired models deactivate with no action from you. Every model lists its live price in sats.
Browse the catalog
- Public page (no account): zapapi.com/models
- API:
GET /ai/models(optionally?type=chat|image|video|embedding|tts|stt)
curl https://api.zapapi.com/api/v1/ai/models \
-H "Authorization: Bearer zap_live_your_key"Each entry includes the model id (used verbatim in requests), display name, type, context length, capability modalities, privacy level, free flag, reasoning config, retirement date, and live sats pricing.
Modalities
| Type | What it does | Priced per |
|---|---|---|
chat | Text conversation (some also emit images) | 1M tokens |
image | Image generation & editing | image / megapixel |
video | Video generation | second / request |
embedding | Vector embeddings | 1M tokens |
tts | Text-to-speech | 1M units |
stt | Speech-to-text | 1M units |
rerank | Document reranking | 1M tokens |
Capability badges
- thinking — reasoning models with selectable effort levels
(
low / medium / high / …). See thinking levels. - vision — accepts image input (
image_urlcontent parts). - image in / audio in / video in — non-text input modalities.
- ZDR — Zero Data Retention endpoints available.
- free — upstream-free variants (
:freesuffix). A small flat platform fee per query applies (free tiers are rate-limited upstream).
Reading prices
Prices are shown in sats per 1M tokens, live at the current BTC/USD rate with the platform fee and FX buffer already included — the number you see is the number you pay.
{
"id": "anthropic/claude-sonnet-5.5",
"type": "chat",
"context_length": 1000000,
"is_free": false,
"reasoning": { "mandatory": true, "supported_efforts": ["max","xhigh","high","medium","low"], "default_effort": "high" },
"pricing_sats": {
"input_per_1M": 61,
"output_per_1M": 305,
"cache_read_per_1M": 3
}
}A typical 1,000-in / 500-out-token query on that model costs roughly
61/1000 + 305/2000 ≈ 0.2 sats at the time of writing. Cheap open models
run a few sats per 1M tokens; frontier models cost more — you always see the
live number before you send.
Variant suffixes
Append a suffix to any model id to steer routing or enable features — no extra configuration:
| Suffix | Effect |
|---|---|
:nitro | Route to the fastest provider |
:floor | Route to the cheapest provider |
:exacto | Prefer strongest tool-calling reliability |
:online | Enable web search for the request |
:thinking | Select the model's reasoning variant |
:extended | Select an extended-context variant |
:free | The free variant of the model (flat platform fee/query) |
:batch | Batch pricing (~50% off, async) |
Free models
Models ending in :free cost nothing upstream but are rate-limited by the
provider. A small flat platform fee per query keeps free usage sustainable
— the fee and remaining free-tier status are visible in
AI → Billing.
Retiring models
Models carry an expiration_date when upstream has announced retirement.
The catalog shows a retiring badge 30 days out, and retired models
deactivate automatically. If you hold a conversation on a retired model, the
API returns model_retired and the web chat restores your message so you can
switch models without losing anything.