PureGamma AIChoose a model, copy the configuration, and ship
A lightweight OpenAI-compatible API Gateway. Use your own sk-pg key and separate prepaid API balance; PureGamma routes only to official provider APIs.
Model catalog
Select a model and the Quick Start, code samples, and parameters update together.
Every model this deployment serves
Read live from the public catalog endpoint. Rows appear here as soon as the catalog serves them, including models without a reviewed summary.
The catalog returned no models.
DeepSeek V4.1 Flash
DeepSeek's current Flash generation, released 10 September 2026. It supersedes the previous V4 Flash build on quality while costing less, and it is the default route for PureGamma chat, agent, report and research workloads. Thinking mode is available but disabled by default on these paths.
deepseek-flash—Pricing details
Catalog quotes are for reference and do not replace billable prices.
Request parameters
Requests follow OpenAI Chat Completions. Provider-specific fields are only forwarded when the selected model supports them.
| Parameter | Type / range | Description |
|---|---|---|
| model | Required | Use the PureGamma model ID shown on this page. |
| messages | Required | A standard OpenAI messages array. |
| stream | Optional | When true, the response uses Server-Sent Events. |
| temperature | 0–2 | Controls randomness; tune it instead of top_p. |
| top_p | 0–1 | Nucleus sampling; normally do not tune it with temperature. |
| max_tokens | 1–131072 | Output ceiling; check the catalog for the actual model limit. |
| response_format | Optional | Check capability support before using JSON mode or JSON Schema. |
| tools / tool_choice | Optional | Tool arguments are untrusted; validate schema and permissions before execution. |
Backend configuration
Keep the key in your server-side secret manager or environment. Vercel, Railway, Cloud Run, Docker, and your own backend only need these three values.
# .env.production / Vercel / Railway / Docker secret store
PUREGAMMA_API_KEY=sk-pg-...
PUREGAMMA_BASE_URL=https://api.puregamma.ai/v1
PUREGAMMA_MODEL=deepseek-flash
# Do not prefix these variables with NEXT_PUBLIC_.
# Do not ship sk-pg keys to the browser, desktop client, or mobile app.The router supports stream, JSON mode, and tools. Check enabled capability with GET /v1/models first; authorize and validate every tool argument on your server before it reaches an external API, payment, database, or file.
const stream = await client.chat.completions.create({
model: "deepseek-flash",
stream: true,
messages: [{ role: "user", content: "Summarize the report." }],
tools: [{
type: "function",
function: {
name: "lookup_price",
description: "Read a price from your own service",
parameters: {
type: "object",
properties: { symbol: { type: "string" } },
required: ["symbol"],
},
},
}],
});The Gateway uses its own prepaid balance. A Stripe-confirmed payment credits the API wallet 1:1 and never consumes a PureGamma product subscription or Credits.