Quickstart
The API serves abliterated open-weight models through OpenAI- and Anthropic-compatible endpoints. Models labelled Official DeepSeek API are the exception: they are DeepSeek's stock models, run by DeepSeek under its own usage and privacy policies (see Models). Any tool or SDK that talks to OpenAI or Anthropic can talk to AbliteratedLLM once you change two settings: the base URL and the API key.
- OpenAI-compatible base URL:
https://api.abliteratedllm.dev/v1 - Anthropic-compatible base URL:
https://api.abliteratedllm.dev
1. Create an account
Open /start, let the browser solve the proof-of-work check, and create an account. You get a 16-digit account number. It is your only login: there is no email and no password. Save it in a password manager now, because it is shown once. If Settings offers a recovery email, you can add one there as a way back in.
2. Add credit
New accounts get $5 of free trial credit when the network qualifies: datacenter, VPN and Tor addresses don't, and each IPv4 /24 or IPv6 /48 gets one trial per 30 days. Trial accounts have tighter limits, listed in Rate limits and trial.
To add credit, open Billing and pay with Monero, Bitcoin or Litecoin. Monero is preselected at checkout; you can switch to Bitcoin or Litecoin on the same page. The minimum top-up is $5, and an account can have up to 3 unpaid invoices open and create up to 10 per hour.
- Credit never expires. It is not refundable.
- If your first credited top-up is $10 or more, you also get a one-time 20% bonus on it, net of network fees: deposit $50, get $60 of credit (one bonus per network every 90 days).
- We credit whatever settles: partial payments, overpayments and late payments inside the payment window.
3. Create an API key
Open API keys and create a key. It starts with sk- and is shown once, so copy it right away. You can give each key a daily, monthly or lifetime spend cap and an expiry date. Details are in Authentication.
4. Send a request
Placeholders:
$API_BASEishttps://api.abliteratedllm.dev/v1,$API_ORIGINishttps://api.abliteratedllm.dev, and$YOUR_KEYis your API key.
In a shell you can run export API_BASE=… and export YOUR_KEY=… once, then paste the curl examples unchanged.
curl
curl $API_BASE/chat/completions \
-H "Authorization: Bearer $YOUR_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "glm-5.3-flash-abliterated",
"messages": [{"role": "user", "content": "Write a two-line poem about the sea."}]
}'
Python
Install the OpenAI SDK with pip install openai, then:
from openai import OpenAI
client = OpenAI(base_url="$API_BASE", api_key="$YOUR_KEY")
stream = client.chat.completions.create(
model="glm-5.3-flash-abliterated",
messages=[{"role": "user", "content": "Write a two-line poem about the sea."}],
stream=True,
)
for chunk in stream:
# The last chunk carries token usage and has an empty choices list.
if chunk.choices:
print(chunk.choices[0].delta.content or "", end="", flush=True)
print()
JavaScript
Install the OpenAI SDK with npm install openai, then:
import OpenAI from "openai";
const client = new OpenAI({ baseURL: "$API_BASE", apiKey: "$YOUR_KEY" });
const stream = await client.chat.completions.create({
model: "glm-5.3-flash-abliterated",
messages: [{ role: "user", content: "Write a two-line poem about the sea." }],
stream: true,
});
for await (const chunk of stream) {
// The last chunk carries token usage and has an empty choices array.
process.stdout.write(chunk.choices[0]?.delta?.content ?? "");
}
Check your balance
curl $API_BASE/balance -H "Authorization: Bearer $YOUR_KEY"
You can also try the models in the browser, without writing code, in the playground.
Models
| Model id | Precision | Input | Cached input | Output | Context | Max output |
|---|---|---|---|---|---|---|
glm-5.3-flash-abliteratedGLM-5.3 Flash Abliterated | FP8 · Zero additional quantization | $0.0977 | $0.0098 | $0.3421 | 262,144 | 65,536 |
deepseek-v4.1-flashDeepSeek V4.1 Flash · Official DeepSeek API | — | $0.30 | $0.006 | $1.20 | 1,048,576 | 393,216 |
deepseek-v4-proDeepSeek V4 Pro · Official DeepSeek API | — | $1.32 | $0.044 | $3.96 | 1,048,576 | 393,216 |
Next steps
- Authentication: headers, keys and spend caps.
- POST /v1/chat/completions, POST /v1/messages and POST /v1/responses: the three generation endpoints.
- Rate limits and trial: request limits, output budgets and what happens when the models are busy.
- Integrations: Claude Code, Codex, Cursor, OpenCode, Pi, Cline, Open WebUI, SillyTavern, Janitor AI, RisuAI and the official SDKs.