Local AI · Uncensored Models
Intelligence,
on your terms.
Open weights. Local inference.
Open-weight refusal adjustment, not prompt injection.
From thought to answer.
01
02
03
One workspace. More intelligence.
Weight-level tuning · Reverse analysis · Authorized security research
Official model ↗
Third-party open-weight variants adjust refusal behavior at the weight level, not through prompt injection. Logos identify base models, not official releases or endorsement. Capabilities depend on deployment and evaluation.
A model lineup for local deployment. Availability depends on the selected weights, license and hardware.
Bring your agents.
Use local models in the development tools you know. Connect to the API of your deployed model.
- 01
Deploy a model
Run inference and a gateway that supports your Agent’s API format.
- 02
Get three values
Your deployment provider supplies the Base URL, API key and model ID.
- 03
Connect your Agent
Add a custom provider, select its protocol and model, then test chat, streaming and tool calls.
API connection details+
Codex
Configure a custom model_provider with base_url, env_key and model. The gateway must support Responses (wire_api = "responses").
Configuration docs ↗Claude Code
Point ANTHROPIC_BASE_URL at a Messages-compatible gateway and configure its key and model mapping. Validate third-party model compatibility separately.
Configuration docs ↗Pi
Configure baseUrl, apiKey, api and model id in models.json. Choose openai-completions or openai-responses to match your gateway.
Configuration docs ↗OpenCode
Set options.baseURL, credentials and models for a custom provider. Use openai-compatible for Chat Completions, or the openai adapter for Responses.
Configuration docs ↗DeepSeek Harness
Use your client version’s custom-model configuration. Verify its gateway protocol, authentication and tool-call format.
Configuration docs ↗Chat Completions request example · Image models use the image endpoint
POST YOUR_GATEWAY_URL/v1/chat/completions
{ "model": "deepseek-v4.1-flash-1017", "messages": […] }Responses / Chat Completions / Anthropic Messages require the matching gateway. Tool use, streaming and model compatibility must be verified per client. No live inference API is hosted by this page.
Open capability. Clear responsibility.
+
Protect children
No child sexual exploitation, abuse, grooming or related content.
Respect authorization
No malicious attacks, unauthorized access, data theft, fraud or harassment. Security research requires permission.
Protect public safety
No drug production or trafficking, violent harm, terrorism or assistance that endangers public safety.
Respect applicable law
Follow the laws applicable to your location and use, across countries and regions, as well as model licenses and third-party terms.