Starveri API Docs

OpenAI-compatible GPT and Codex access with credits, credit-aware usage modes, and documented model pricing.

Pricing

Model prices are in credits per 1M tokens and are advertised at documented model pricing. The example column shows the approximate price for a request with 10,000 input tokens and 10,000 output tokens before any tool fees. Requests using more than 30% of a model's context window use the long-context multiplier. Gemma models do not offer input-token caching or cached-input discounts.

Model	Cached Input / 1M	Input / 1M	Output / 1M	Example 10k in + 10k out
Gemma 4 `gemma-4`	-	5 credits	15 credits	0.2 credits
Gemma 4 Uncensored `gemma-4-uncensored`	-	10 credits	30 credits	0.4 credits
GPT 5.6 Sol `gpt-5.6-sol`	5 credits	50 credits	150 credits	2 credits
GPT 5.6 Terra `gpt-5.6-terra`	2.5 credits	25 credits	75 credits	1 credit
GPT 5.6 Luna `gpt-5.6-luna`	1 credit	10 credits	30 credits	0.4 credits
GPT 5.5 `gpt-5.5`	5 credits	50 credits	150 credits	2 credits
GPT 5.4 `gpt-5.4`	1.67 credits	16.67 credits	50 credits	0.6667 credits
GPT 5.4 Mini `gpt-5.4-mini`	0.56 credits	5.56 credits	16.67 credits	0.2223 credits
GPT 5.3 Codex Spark `gpt-5.3-codex-spark`	2 credits	20 credits	60 credits	0.8 credits

Tool	Price
Web search	0.1 credits per call
Code interpreter	0.5 credits per GB RAM/minute plus 0.5 credits per GB disk/minute (minimum 1 GB RAM and 5 GB disk)

Quick Start

Use the credits page to buy at least $1 in prepaid credits.
After checkout, copy your dashboard API key or create a child API key from the dashboard.
Send OpenAI-compatible requests to https://api.starveri.net/v1.

curl https://api.starveri.net/v1/chat/completions \
  -H "Authorization: Bearer YOUR_STARVERI_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"gpt-5.5","messages":[{"role":"user","content":"Say hello"}]}'

Free Demo

The credits page includes a GPT 5.5 no-reasoning demo for quick text-only testing before purchase.

5 demo requests per IP.
2,000 total demo tokens per IP.
1 concurrent demo request per IP.
No files, images, web search, or code interpreter in demo mode.

Models

Use GET /v1/models for the live supported model list and prices. Unsupported models are not advertised.

Model	Notes
`gpt-5.6-sol`	GPT 5.6 Sol; Standard service, Pro, and Max reasoning are available.
`gpt-5.6-terra`	GPT 5.6 Terra; Standard service, Pro, and Max reasoning are available.
`gpt-5.6-luna`	GPT 5.6 Luna; Standard service and Pro are available; Pro keeps Luna's normal effort mapping.
`gpt-5.5`	GPT 5.5 with None through Xhigh reasoning, Standard service, Flex, and Pro where supported.
`gpt-5.4`	GPT 5.4 with None through Xhigh reasoning and Standard service.
`gpt-5.4-mini`	GPT 5.4 Mini with Standard and Pro Standard service.
`gpt-5.3-codex-spark`	Text-only Spark. Image inputs are first transcribed by GPT 5.4 Mini and both stages are billed.
`gemma-4`	Regular Gemma 4 served by a dedicated runtime. Standard inference only, 124,160-token context, automatic request/response moderation, and a binary `thinking` control.
`gemma-4-uncensored`	Gemma 4 Uncensored served by a dedicated runtime. Standard inference only, 32,000-token context, no automatic omni or policy moderation, and a binary `thinking` control; account and rate controls still apply.

Credits and access modes

Purchases add prepaid credits immediately. The current model table and response metadata are the source of truth for token and tool pricing; availability and attachment handling can vary by route. Gemma requests use a dedicated runtime only. During capacity pressure, requests may wait in arrival order; unavailable dedicated inference returns HTTP 503 rather than switching models.

Use the dashboard to issue child keys, assign balances, grant, set, or remove credits, and revoke keys. No fixed per-key quota table is published here.

Token Buckets

For marketplaces that bill per request, Starveri exposes bucketed endpoints that cap the request size and bill from the selected bucket.

POST /v1/chat/completions/gpt-5.5-10k
POST /v1/chat/completions/gpt-5.5-100k
POST /v1/chat/completions/gpt-5.4-mini-10k

Credits

Credits are prepaid balance. The dashboard can issue child keys, assign credit balances, manage child-key balances and credit operations, restock keys, and revoke keys.

Stripe Checkout supports coupon codes when available.

Tools

Paid API requests may use web search and code interpreter when available for the route. The code interpreter is sandboxed with timeout, input-size, output-size, and concurrency limits.