Open-weight models / Chat & API

Uncensored AI API.
Choose the model.

Use GLM-5.3 Raw through an OpenAI-compatible API, or explore Antislop for writing and conversation. Start in chat, compare answers on your own tasks, and connect the model that fits your workflow.

Get an API key ↗Compare models ↗

Models and API pricing

Chat models are billed by token against a prepaid balance. Input and output have separate rates. The table uses the same live registry as the model catalogue; open a model for its output limit and access details.

RawModels chat model prices in USD per million tokens
Model	Input / 1M	Output / 1M	Context	Tools	Access
GLM-5.3 Raw	$2.50	$7.00	1M	Yes	Available
Qwen3.8 27B Antislop	$0.60	$5.00	262.1K	Yes	Available

A long conversation sends its history as input on each turn. Your context length, generated output and repeated attempts all affect cost. Check the model's current price and set an output budget before running a larger workload.

Connect an existing API client

Sign in, create a key in API key settings and fund your balance when needed. Set your client's base URL to https://www.rawmodels.ai/v1, supply the key, and use the model ID from the catalogue. The chat completions endpoint supports streamed responses.

cURL

curl https://www.rawmodels.ai/v1/chat/completions \
  -H "Authorization: Bearer $RAWMODELS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "glm-5.3-raw",
    "messages": [{"role": "user", "content": "Write a short opening scene for a mystery story."}],
    "stream": true,
    "max_tokens": 512
  }'

Python example, parameters and error reference ↗

Use your own interface

Writing and roleplay

Keep your character cards, story and settings in a compatible frontend. Our SillyTavern setup guide covers the base URL, model selection, a first test and common connection errors. For the writing fine-tune, read the Antislop training and release notes.

Apps and coding tools

Use the API directly from your application. For a tool-using workflow, choose a model whose catalogue entry lists tool calling. The Claude Code guide explains model mapping and configuration; the API reference describes supported parameters.

Before you choose a model

What does “uncensored LLM” mean here?: The term describes models offered for more direct responses and fewer blanket refusals. Behavior depends on the checkpoint and the request; it is not a guarantee that every prompt will be answered. Antislop's writing focus is a separate characteristic. The service's acceptable-use terms still apply.
Can I try the models without writing code?: Yes. Open an available chat model from the catalogue and use the web chat after signing in. Chat and API usage draw from the same account balance.
Are these the same models as OpenAI's?: No. OpenAI-compatible describes the API format. The selected RawModels model generates the answer. Its training, capabilities and limitations differ from other models.
Does “uncensored” mean private or zero retention?: No. Model behavior and data handling are separate. Review the current data-storage terms before sending content.
Can OpenJev be used as a chat model?: OpenJev returns typed decisions rather than generated chat. Use the separate Decisions API for choices, yes/no answers and scores over text and images.

Uncensored AI API.Choose the model.