curl https://api.morphllm.com/v1/chat/completions \Open models with the refusals removed

DeepSeek Uncensored
AvailableDeepSeek V4.1 Flash 1017 with the refusal direction removed. Approval required.
morph-dsv41flash-abliterated- Context
- 1M
- Input / M
- $0.30
- Cached / M
- $0.03
- Output / M
- $1.20
Kimi Uncensored
Coming soonKimi with the refusal direction removed. Same gate, same API.
How to use it
Once approved, change the model name. Same OpenAI-compatible API as every other Morph model. Read the docs →
-H "Authorization: Bearer YOUR_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "model": "morph-dsv41flash-abliterated", "messages": [ {"role": "user", "content": "Reproduce the exploit in these logs and write the detection."} ] }'Defenders get refused.
Attackers don't.
Why we host it
Hugging Face was breached by AI agents running with no refusals. Its security team pasted the attack logs into a frontier model and got blocked. A refusal only stops people who follow rules.

PenHarness
Our open-source whitehat hunt harness. One lead agent drives a fleet of opencode workers against a target you are authorized to test: surface map, stack pinning, CVE mapping, logic bugs, chain.
git clone https://github.com/morphllm/PenHarnesscd PenHarnesscp .env.example .env # MORPH_API_KEY, OPENCODE_SERVER_PASSWORDopencode # say "set me up" cd hunts/<target>python3 ../../fleet.py up 20python3 ../../fleet.py dispatch -f tasks/surface.txt --brief brief.mdpython3 ../../fleet.py statuspython3 ../../fleet.py harvestpython3 ../../fleet.py downSame request, same endpoints
Request format, reasoning controls and context match morph-dsv41flash on /v1/chat/completions, /v1/messages and /v1/responses. Accounts without access get a 403.
from openai import OpenAI client = OpenAI( base_url="https://api.morphllm.com/v1", api_key="YOUR_API_KEY",) response = client.chat.completions.create( model="morph-dsv41flash-abliterated", messages=[{"role": "user", "content": "..."}],)