A DeepSeek API with no logging

What “no logging” means for this model

• We store no prompts or replies. One row of counts per call, so we can bill you: time, key, model, tokens, credits and status, kept for 30 days.
• The host stores none either. DeepSeek runs on a private tier that stores no prompts or replies, and it receives calls from us, not from you.
• Nothing is used for training. There is nothing kept to train on.
• No account details. A key needs no email or name; you buy credits and get an account key.

Every piece we keep, field by field, is on What we keep. How that compares with typical API data policies is on the private LLM API page.

Pin the 1017 build

Pictures

DeepSeek reads up to 10 pictures per request as image_url parts: a link, or a data: URL for a file on your machine. Screenshots, receipts, charts, photos. A picture is billed as the input tokens the model reads it as, with no separate charge. The picture is read and dropped like the rest of the prompt.

curl https://nokeep.ai/v1/chat/completions \
  -H "Authorization: Bearer $NOKEEP_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "deepseek-v4.1-flash-1017-uncensored",
  "messages": [
    {
      "role": "user",
      "content": [
        {
          "type": "image_url",
          "image_url": {
            "url": "https://example.com/screenshot.png"
          }
        },
        {
          "type": "text",
          "text": "What is wrong in this stack trace?"
        }
      ]
    }
  ]
}'

More on sizes, data URLs and limits: Pictures.

Thinking

DeepSeek answers at once unless you ask it to think. Set reasoning_effort to low, high or max for planning, maths and code; leave it out, or send none, for a fast reply. The thinking comes back apart from the answer, in message.reasoning_content, and is billed as output tokens.

import os
from openai import OpenAI

client = OpenAI(
    base_url="https://nokeep.ai/v1",
    api_key=os.environ["NOKEEP_API_KEY"],
)

reply = client.chat.completions.create(
    model="deepseek-v4.1-flash-1017-uncensored",
    reasoning_effort="high",
    max_tokens=4000,
    messages=[{"role": "user", "content": "Find the race condition in this function: ..."}],
)
print(reply.choices[0].message.reasoning_content)
print(reply.choices[0].message.content)

With thinking on, give the reply 1,000 tokens or more. Under that, the model can spend them all thinking and return an empty answer, which is still billed. See Thinking.

Price

$0.45 in · $0.011 cached · $1.75 out per 1M tokens. A long chat or an agent resends its history, and the part the model read on a recent call bills at the cached price, so most of a long session’s input costs a small fraction of the full rate. Paid from prepaid credits: 1,000 are $1, from $5, and they never expire.