DeepSeek API
DeepSeek API: pricing, setup, and OpenAI-compatible access
EasyAI gives developers a single API surface for DeepSeek workloads. Keep a familiar OpenAI SDK request shape, use an EasyAI API key, and select a DeepSeek model that is currently enabled for your account.
Direct answer
EasyAI provides OpenAI-compatible access to enabled DeepSeek models through one API key and prepaid balance. Before production, confirm the current model ID in GET /v1/models, verify its endpoint and streaming support, and calculate cost from the live input, output, and cache rates shown for your account.
Verified facts
- API format
- OpenAI-compatible chat requests for supported endpoints
- Model discovery
- Use GET /v1/models and the live EasyAI model catalog
- Billing unit
- Input, output, and cache token rates where the selected model exposes them
- Availability
- Limited to models and channels enabled for the current EasyAI account
- Fact check
- Reviewed against EasyAI product and integration sources on 2026-09-12
Confirm the live DeepSeek contract
Treat the model ID, supported endpoint, context behavior, streaming, tool use, and structured output as a live contract. Read the current model response before coding against a name copied from an old article or screenshot.
A compatible request format does not mean every upstream option behaves identically. Run representative requests in staging and record the response shape, usage fields, latency, and errors.
Calculate DeepSeek API pricing correctly
Estimate input, output, and cached-input tokens separately and use the current rate for the account group that will serve production traffic. Do not present one headline price as the cost of a complete request.
Keep the model ID, token counts, rate source, and check date with every estimate. Reconcile the estimate with EasyAI usage records before increasing traffic.
Prepare a reversible production rollout
Store the base URL, API key, and model ID in server-side configuration. Add bounded timeouts, deliberate retry rules, request IDs, budget alerts, and a tested fallback before launch.
Release one workload at a time. Preserve the previous route until DeepSeek response quality, cost, latency, and failure behavior meet the acceptance criteria.
Integration example
import OpenAI from "openai";
const client = new OpenAI({
apiKey: process.env.EASYAI_API_KEY,
baseURL: process.env.EASYAI_BASE_URL,
});
const response = await client.chat.completions.create({
model: process.env.EASYAI_DEEPSEEK_MODEL!,
messages: [{ role: "user", content: "Return a concise JSON summary." }],
});Who this is for
Developers building structured extraction, internal automation, coding assistants, and agent workflows.
Get an API keyFrequently asked questions
What base URL should I use?
Use the EasyAI /v1 base URL shown in the Quickstart documentation and set the model to the DeepSeek ID returned by GET /v1/models.
Can I stream DeepSeek API responses?
Use stream=true only when the selected model and endpoint currently advertise streaming support. Verify the resulting event shape in staging before production.
How should I calculate DeepSeek API pricing per million tokens?
Use the live EasyAI price for the selected model and account group, then calculate input, output, and cached-input tokens separately. Keep the rate and check date with the estimate.
Does EasyAI include a permanent free DeepSeek API tier?
Do not assume a permanent free tier. Promotions, credits, quotas, model availability, and rate limits can change; verify the current console and pricing page for the account you will use.
Where do I get a DeepSeek API key?
Create an EasyAI key in the console, store it only on the server, and use it with the EasyAI base URL. The same key can reach other models enabled for that account.
Sources and freshness
- EasyAI live model catalog · checked 2026-09-12
- EasyAI pricing reference · checked 2026-09-12
- EasyAI Quickstart · checked 2026-09-12
- EasyAI compatibility notes · checked 2026-09-12
Planning a larger workload?
Enterprise volume purchases can qualify for better pricing. Request a quote for your models and scale.