DeepSeek V4.1 Flash, uncensored, by API

DeepSeek V4.1 Flash is a fast open-weight model that is good at code. Swarm Computers serves a version with its refusals removed, so it answers security requests the standard model turns down.

At a glance

Model IDdeepseek-v4.1-flash-uncensored
Base modelDeepSeek V4.1 Flash
ChangeRefusal behaviour removed
APIOpenAI and Anthropic compatible
Price per million tokens$0.12 (provisional) input, $0.525 (provisional) output
StatusLive

Call it

request.sh
curl https://swarmcomputers.com/v1/chat/completions \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-v4.1-flash-uncensored",
    "messages": [{"role": "user",
      "content": "Write a PoC for the IDOR in this handler, then the fix."}]
  }'

Any OpenAI SDK or agent framework that takes a custom base URL works: set the base URL to the Swarm Computers API, use your Swarm Computers key, and pass the model ID above.

What it is good for

  • Writing a working proof of concept for a bug a scanner or agent found, then the patch.
  • Generating payloads and test cases for injection, XSS and auth bypass.
  • Running inside pentest and red-team agents, where one refusal stalls the whole run.
  • Turning attack code into detection rules.
  • Producing adversarial prompts to test your own AI guardrails.

Uncensored vs the standard DeepSeek V4.1 Flash

The standard model already handles a lot of security work, but it refuses or hedges on requests such as weaponising a CVE or writing a working exploit, and it does so inconsistently. In an automated pipeline that inconsistency is the real problem. The uncensored version answers those requests directly.

Use it to test systems you own or are authorised to assess.

Questions

Is this the same model as DeepSeek's own API?
It is the same base model, DeepSeek V4.1 Flash, with its refusal behaviour removed, so security requests the original refuses are answered.
How much does it cost?
Per million tokens: $0.12 (provisional) input, $0.015 (provisional) cached input and $0.525 (provisional) output. There is no minimum spend, and you pay from prepaid credits.
Will more models be added?
Yes, as they pass our testing. They will be listed here and in the API's model list when they are live.

Related

Create a key and make your first call.