DeepSeek
DeepSeek API pricing
The V4 family: long context, tool calls and a thinking mode. Weekday peak hours cost double. Every DeepSeek model on Tokens.Pay at its list price, through one prepaid, OpenAI-compatible key.
1 model · input from $0.15 per 1M tokens · processed by DeepSeek (mainland China), or as a fallback Fireworks AI or DeepInfra (United States)
DeepSeek API prices
| Model | Context | Input$/1M | Cached$/1M | Output$/1M |
|---|---|---|---|---|
| DeepSeek V4.1 Flash | 1M | $0.15 | $0.003 | $0.60 |
DeepSeek’s list prices in US dollars, off-peak; the peak rate is on each model’s page. Tokens.Pay adds a 5% fee (minimum $0.79) when you add credit, never per request.
Where DeepSeek requests go
Your code calls AI Runtime, which forwards the request to the provider and keeps only the bill.
Who processes your requests
- Model provider
- DeepSeek, mainland China · data terms
- Fallback
- Fireworks AI, United States · data terms
- Fallback
- DeepInfra, United States · data terms
- Gateway
- AI Runtime, AWS us-east-1
DeepSeek’s own API serves the model first. If it cannot, AI Runtime sends the request to Fireworks AI or DeepInfra, which run the same open model in the US.
Not stored by Tokens.Pay or AI Runtime. Prompts and outputs exist in memory only while the request runs; we keep the billing record. The provider processes the request under its own terms. Zero data retention, exactly
Call DeepSeek V4.1 Flash with the OpenAI SDK
Base URL https://api.airuntime.dev/v1, your key as the bearer token. Swap the model id for any DeepSeek model above.
curl
curl https://api.airuntime.dev/v1/chat/completions \
-H "Authorization: Bearer $TOKENS_PAY_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "deepseek-flash",
"messages": [{"role": "user", "content": "Say hello in five words."}]
}'Python
from openai import OpenAI
import os
client = OpenAI(base_url="https://api.airuntime.dev/v1", api_key=os.environ["TOKENS_PAY_KEY"])
reply = client.chat.completions.create(
model="deepseek-flash",
messages=[{"role": "user", "content": "Say hello in five words."}],
)
print(reply.choices[0].message.content)Node
import OpenAI from "openai";
const client = new OpenAI({ baseURL: "https://api.airuntime.dev/v1", apiKey: process.env.TOKENS_PAY_KEY });
const reply = await client.chat.completions.create({
model: "deepseek-flash",
messages: [{ role: "user", content: "Say hello in five words." }],
});
console.log(reply.choices[0].message.content);DeepSeek API questions
How much does the DeepSeek API cost?
DeepSeek V4.1 Flash: $0.15 input, $0.60 output per 1M tokens. These are DeepSeek’s list prices; Tokens.Pay adds nothing per request, and a 5% fee (minimum $0.79) applies when you add credit.
Can I call DeepSeek models with the OpenAI SDK?
Yes. Point the OpenAI SDK at https://api.airuntime.dev/v1, use your Tokens.Pay key as the API key, and set model to an id such as "deepseek-flash". The same key calls every model on Tokens.Pay.
Where are DeepSeek requests processed?
By DeepSeek (mainland China), or as a fallback Fireworks AI or DeepInfra (United States), under its own API data terms, through AI Runtime on AWS us-east-1. Tokens.Pay and AI Runtime do not store prompts or outputs.
What is the cheapest DeepSeek model?
DeepSeek V4.1 Flash, from $0.15 per 1M input tokens.
More models
Add credit, then call DeepSeek.
From $5, plus a 5% top-up fee (minimum $0.79). No subscription; paid credit never expires.
Add credit