How to Get a DeepSeek API Key (Plus a Flat-Rate Alternative)

Getting a DeepSeek API key takes three steps on the official platform: create an account, open the API keys page, and generate a key. The part most guides skip is what happens after: as of August 16, 2026, DeepSeek’s API now charges peak and off-peak prices that differ by 2x, and there’s a same-model alternative that skips the identity verification step and bills a flat rate instead.

DeepSeek’s own price now shifts by time of day. Novita AI serves the same DeepSeek-V4-Flash-0731 and DeepSeek-V4-Pro-0813 models at flat rates that don’t:

Model	DeepSeek Official (off-peak / peak)	Novita AI (flat rate)
DeepSeek-V4-Flash-0731	Input $0.22 / $0.44 · Output $0.66 / $1.32	Input $0.44 · Output $1.32
DeepSeek-V4-Pro-0813	Input $0.66 / $1.32 · Output $1.98 / $3.96	Input $1.32 · Output $3.96

Prices are per 1M tokens. Source: DeepSeek pricing and Novita AI pricing, both checked 2026-08-24. Novita’s flat rate matches DeepSeek’s own peak rate for both models — it isn’t a discount on DeepSeek’s off-peak price. What it buys you is a price that doesn’t change by time of day, plus a signup flow without identity verification — details below.

Key Takeaways

• DeepSeek’s official API key process takes under 5 minutes: sign up, verify a phone number or email, generate the key from the platform console.
• DeepSeek switched to peak/off-peak pricing on August 16, 2026. Peak hours (01:00–04:00 and 06:00–10:00 UTC) cost double the off-peak rate for both DeepSeek-V4-Flash-0731 and DeepSeek-V4-Pro-0813.
• Novita AI serves DeepSeek-V4-Flash (model ID deepseek/deepseek-v4-flash-0731) at a flat $0.44/M input and $1.32/M output — the same as DeepSeek’s own peak rate, not a discount on the off-peak price.
• Novita AI serves DeepSeek-V4-Pro (model ID deepseek/deepseek-v4-pro-0813) at a flat $1.32/M input and $3.96/M output — also matching DeepSeek’s peak rate, not the off-peak rate.
• Neither model is cheaper on Novita than running it on DeepSeek’s own API during off-peak hours. The trade is flat, predictable billing instead of a discount.
• Novita’s API key setup skips DeepSeek’s identity verification and works with the same OpenAI-compatible request format.

How Much Does a DeepSeek API Key Cost to Get?

Creating the key itself is free. What costs money is the token usage once you start calling the model, and that’s where the August 2026 pricing change matters. DeepSeek’s model pricing now depends on the time of day you send the request, not just the token count.

DeepSeek’s New Peak/Off-Peak Pricing (Effective August 16, 2026)

DeepSeek confirmed the pricing change in its API change log: starting at 16:00 UTC on August 16, 2026, DeepSeek-V4-Flash-0731 and DeepSeek-V4-Pro-0813 bill at different rates depending on the hour. Off-peak prices are exactly half the peak rate. Peak hours are 01:00–04:00 UTC and 06:00–10:00 UTC, Monday through Friday; every other hour is off-peak.

Model	Metric	Off-peak	Peak
DeepSeek-V4-Flash-0731	Input (cache miss), per 1M tokens	$0.22	$0.44
DeepSeek-V4-Flash-0731	Output, per 1M tokens	$0.66	$1.32
DeepSeek-V4-Pro-0813	Input (cache miss), per 1M tokens	$0.66	$1.32
DeepSeek-V4-Pro-0813	Output, per 1M tokens	$1.98	$3.96

Source: DeepSeek API pricing page, checked 2026-08-24. Quoted from the page: “Off-peak rates are half of the peak rates. Peak hours are 01:00 - 04:00 and 06:00 - 10:00 UTC, Monday through Friday (all other hours are off-peak).”

If your traffic happens to land inside those two peak windows — which cover 7 of 24 hours a day — your bill for the same workload roughly doubles compared to running it off-peak. That’s a scheduling problem most teams didn’t have to think about before August 16.

DeepSeek V4-Flash and V4-Pro Pricing: DeepSeek vs. Novita AI

Here’s the same comparison from the top of this article broken out by peak and off-peak rate, so you can see exactly which DeepSeek window each Novita price matches. Both DeepSeek-V4-Flash-0731 and DeepSeek-V4-Pro-0813 are also served through Novita AI’s API at a flat rate that doesn’t change by time of day.

Provider	Model	Input ($/M)	Output ($/M)
Novita AI	DeepSeek-V4-Flash-0731	$0.44	$1.32
DeepSeek (off-peak)	DeepSeek-V4-Flash-0731	$0.22	$0.66
DeepSeek (peak)	DeepSeek-V4-Flash-0731	$0.44	$1.32
Novita AI	DeepSeek-V4-Pro-0813	$1.32	$3.96
DeepSeek (off-peak)	DeepSeek-V4-Pro-0813	$0.66	$1.98
DeepSeek (peak)	DeepSeek-V4-Pro-0813	$1.32	$3.96

Source: Novita AI pricing page, checked 2026-08-24. Quoted from the page: model “Deepseek V4 Flash 0731” lists Input “$0.44 /Mt” and Output “$1.32 /Mt”; model “DeepSeek V4 Pro 0813” lists Input “$1.32 /Mt” and Output “$3.96 /Mt”.

Read the two rows carefully, because both models land in the same place:

• V4-Flash-0731: Novita’s flat $0.44/$1.32 matches DeepSeek’s own peak rate exactly. It’s not cheaper than DeepSeek’s off-peak rate ($0.22/$0.66) — if you only ever called V4-Flash during off-peak hours, DeepSeek’s own API would cost less.
• V4-Pro-0813: Same pattern. Novita’s price ($1.32/$3.96) is identical to DeepSeek’s own peak rate, not a discount on it.

Both providers serve the same model weights, so neither Novita rate beats DeepSeek’s best-case (off-peak) price on either model. What Novita buys you instead is a rate that never changes: you pay the peak-equivalent price at 3am or 3pm alike, with no need to track UTC hours or hold traffic until the window opens.

That trade only makes sense for some workloads. If you already schedule requests to land inside DeepSeek’s off-peak windows, DeepSeek’s own API is cheaper than Novita’s flat rate for both models — there’s no reason to switch on price. Novita’s advantage is operational: no peak/off-peak logic to build, and no identity verification step to get a key. And if your team needs to create or delete keys programmatically through an API rather than a console, neither provider supports that today (see the FAQ below), so this guide won’t solve that specific problem.

Step 1: Get Your Novita API Key

Novita AI’s key creation flow requires only an account, not a separate identity check:

1. Sign in or create an account at novita.ai.
2. Open Key Management in the console.
3. Select Create API Key, name it something you’ll recognize later (like production or local-test), and copy the key.

The full key is shown only once, at creation. Every key starts with the sk_ prefix, stays valid indefinitely once created, and each account can hold up to 10 keys. The Novita API does not expose an endpoint to create or delete keys — that part stays console-only, so budget a minute for the console step even if the rest of your setup is scripted.

Set it as an environment variable before writing any code:

export NOVITA_API_KEY="your_api_key_here"

Step 2: Confirm the Model ID and Endpoint

Field	Value
Model ID (Flash)	deepseek/deepseek-v4-flash-0731
Model ID (Pro)	deepseek/deepseek-v4-pro-0813
Base URL	https://api.novita.ai/openai
Chat completions endpoint	https://api.novita.ai/openai/v1/chat/completions
Context window	1,048,576 tokens
Input pricing (Flash)	$0.44 / M tokens
Output pricing (Flash)	$1.32 / M tokens

Novita also lists an unsuffixed deepseek/deepseek-v4-flash and deepseek/deepseek-v4-pro model ID, each priced differently from the -0731/-0813 versions above. Use the dated suffix — it’s the current, actively promoted entry in Novita’s model catalog. Check the DeepSeek-V4-Flash-0731 model page before shipping — pricing and limits can change, and this table is a snapshot checked 2026-08-24.

Step 3: Send Your First Request

import os
from openai import OpenAI

client = OpenAI(
    base_url="https://api.novita.ai/openai",
    api_key=os.environ["NOVITA_API_KEY"],
)

response = client.chat.completions.create(
    model="deepseek/deepseek-v4-flash-0731",
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "Explain what a rate limiter does in one sentence."},
    ],
    max_tokens=200,
)

print(response.choices[0].message.content)

curl "https://api.novita.ai/openai/v1/chat/completions" \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $NOVITA_API_KEY" \
  -d '{
    "model": "deepseek/deepseek-v4-flash-0731",
    "messages": [{"role": "user", "content": "Explain what a rate limiter does in one sentence."}],
    "max_tokens": 200
  }'

The API is OpenAI-compatible, so if you already have an OpenAI client set up, changing base_url and model is the only edit needed.

Step 4: Read the Response and Check Your Balance

The response text lands at response.choices[0].message.content, same as any OpenAI-compatible chat completion. Track usage.prompt_tokens and usage.completion_tokens from the response if you want to reconcile spend against the pricing table above — since Novita’s rate is flat, that reconciliation doesn’t require knowing what time your request went out, which is one less variable than tracking DeepSeek’s own peak/off-peak billing.

FAQ

How long does it take to get a DeepSeek API key?

A few minutes on either platform. DeepSeek’s official signup includes an identity verification step (phone or email); Novita’s console skips that and only requires an account.

Does the DeepSeek API key work the same on Novita AI?

No — they’re separate keys from separate providers. A DeepSeek API key authenticates against api.deepseek.com; a Novita API key authenticates against api.novita.ai. Both serve the same DeepSeek-V4-Flash-0731 and DeepSeek-V4-Pro-0813 model weights through an OpenAI-compatible format, but you need the matching key for whichever base URL you call.

Is DeepSeek-V4-Flash actually cheaper on Novita AI?

No, not on price alone. Novita’s flat $0.44/$1.32 per million tokens for V4-Flash-0731 matches DeepSeek’s own peak rate — it’s higher than DeepSeek’s off-peak rate of $0.22/$0.66. The same is true for V4-Pro-0813: Novita’s price matches DeepSeek’s peak rate, not a discount on it. What Novita offers instead is a rate that doesn’t change by time of day and a signup flow without identity verification.

What are DeepSeek’s peak hours?

01:00–04:00 UTC and 06:00–10:00 UTC, Monday through Friday. All other hours are off-peak, billed at half the peak rate. This schedule and the peak/off-peak split took effect August 16, 2026.

Can I create or delete a DeepSeek API key through the API instead of the console?

No, on either platform. Both DeepSeek’s and Novita’s official APIs cover listing keys and managing model-access policy, but key creation and deletion stay console-only actions.

Recommended Articles

• DeepSeek-V4-Flash on Novita AI: Fast Reasoning at Lower Cost
• DeepSeek V4 Pro 0813 on Novita AI: GA Release, DSpark, and Pricing
• Simple Guide to Using LLM API Key