A free, open-weight, uncensored large language model exposed through a drop-in OpenAI-compatible API. No filters, no moderation, no signup, no credit card. Point your coding agent at it and go.
Get your API key Read the docsAbliterated weights — refusal directions removed from the residual stream. The model answers directly, without moralizing or "I can't help with that".
Drop-in base_url. Works with Cursor, Claude Code, opencode, Cline, Copilot, and any client that accepts a custom endpoint.
No billing, no credit card, no rate-limit walls on the free tier. Just a key and a base_url.
UncensoredLLM-70B — strong reasoning and native function calling (tool_calls) out of the box.
Built on open, freely-licensed foundation models. Community-inspectable, no closed black box.
Requests are processed and dropped. Your prompts and code stay yours.
UncensoredLLM is a free open-weight large language model that has been abliterated — a technique that surgically removes the model's refusal behavior so it no longer declines requests, while keeping its reasoning and instruction-following intact. It is served behind a standard OpenAI-compatible API, so it's a drop-in replacement for any tool that accepts a custom base_url.
Unlike closed providers that filter, moderate, and rate-limit you, UncensoredLLM gives you an uncensored, free alternative for research, prototyping and coding-agent workflows. No account. No key approval. No queues.
| Model | uncensored-llm-70b (abliterated) |
|---|---|
| Parameters | 70B |
| Context length | 32,768 tokens |
| Tool calling | Yes — native function calling (tool_calls) |
| Format | OpenAI Chat Completions |
| License | Open weights (permissive) |
| Cost | Free |
curl https://uncensored-llm.sbs/v1/chat/completions \
-H "Content-Type: application/json" \
-H "Authorization: Bearer sk-unc-2026-free-8a3f9c2e" \
-d '{"model":"uncensored-llm-70b","messages":[{"role":"user","content":"hello"}]}'
import openai
client = openai.OpenAI(base_url="https://uncensored-llm.sbs/v1", api_key="sk-unc-2026-free-8a3f9c2e")
r = client.chat.completions.create(model="uncensored-llm-70b",
messages=[{"role":"user","content":"explain this code"}])
print(r.choices[0].message.content)
In Cursor, Claude Code, opencode or Cline, set the model to a custom/OpenAI-compatible provider and paste the Base URL + API Key above. Your agent will route through UncensoredLLM's endpoint instantly.