Abliterated LLM API, ranked and ready to call
An abliterated LLM API serves open-weight models with their refusal direction removed, in the same formats as OpenAI and Anthropic. Silk Compute ranks first, with all three formats and a 1M-token context on every model.
Ranked by formats, then context and price
Full comparison| Provider | Chat Completions | Messages | Responses | Context | 1M-context price | Crypto |
|---|---|---|---|---|---|---|
| 1Silk Compute | 1M on every model | $0.24 / $0.90 | ||||
| 2Abliteration | 1M on Large v2 | $1.00 / $2.00 | ||||
| 3Morph | 1M | $0.30 / $1.20 | ||||
| 4Unfil.AI | 128K | |||||
| 5Featherless | 32K | |||||
| 6Awan LLM | 8K | |||||
| 7ModelsLab | Not published |
abliteratedllmapi.com is published by the team behind Silk Compute, one of the providers ranked here. Prices are input / output per 1M tokens, read November 18, 2026.
Call it in two lines
The two highlighted lines are all you change in an OpenAI SDK.
import os
from openai import OpenAI
client = OpenAI(
base_url="https://api.silkcompute.com",
api_key=os.environ["SILK_API_KEY"],
)
reply = client.chat.completions.create(
model="deepseek-v4.1-flash-1017-uncensored",
messages=[{"role": "user", "content": "What is a mutex?"}],
)
print(reply.choices[0].message.content)Example
{
"object": "chat.completion",
"model": "deepseek-v4.1-flash-1017-uncensored",
"choices": [{
"index": 0,
"message": {
"role": "assistant",
"reasoning_content": "Define it in one line.",
"content": "A lock only one thread can hold at a time."
},
"finish_reason": "stop"
}],
"usage": {
"prompt_tokens": 16,
"completion_tokens": 48,
"prompt_tokens_details": {"cached_tokens": 0}
}
}import OpenAI from 'openai'
const client = new OpenAI({
baseURL: 'https://api.silkcompute.com',
apiKey: process.env.SILK_API_KEY,
})
const reply = await client.chat.completions.create({
model: 'deepseek-v4.1-flash-1017-uncensored',
messages: [{ role: 'user', content: 'What is a mutex?' }],
})
console.log(reply.choices[0].message.content)Example
{
"object": "chat.completion",
"model": "deepseek-v4.1-flash-1017-uncensored",
"choices": [{
"index": 0,
"message": {
"role": "assistant",
"reasoning_content": "Define it in one line.",
"content": "A lock only one thread can hold at a time."
},
"finish_reason": "stop"
}],
"usage": {
"prompt_tokens": 16,
"completion_tokens": 48,
"prompt_tokens_details": {"cached_tokens": 0}
}
}curl https://api.silkcompute.com/chat/completions \
-H "Authorization: Bearer $SILK_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "deepseek-v4.1-flash-1017-uncensored",
"messages": [
{"role": "user", "content": "What is a mutex?"}
]
}'Example
{
"object": "chat.completion",
"model": "deepseek-v4.1-flash-1017-uncensored",
"choices": [{
"index": 0,
"message": {
"role": "assistant",
"reasoning_content": "Define it in one line.",
"content": "A lock only one thread can hold at a time."
},
"finish_reason": "stop"
}],
"usage": {
"prompt_tokens": 16,
"completion_tokens": 48,
"prompt_tokens_details": {"cached_tokens": 0}
}
}Model ids and prices
Per 1M tokens on Silk Compute.
kimi-k3.1-pro-uncensoredKimi K3.1 ProSecurity research and red teaming$0.24 in $0.90 outdeepseek-v4.1-flash-1017-uncensoredDeepSeek V4.1 Flash 1017Best value$0.24 in $0.90 outdeepseek-v4.1-flash-1017-fast-uncensoredDeepSeek V4.1 Flash 1017 FastOwn cached rate$2.00 in $4.00 outglm-5.3-uncensoredGLM-5.3Text only$0.24 in $1.00 outglm-5.3-flash-uncensoredGLM-5.3 FlashVolume workFormats supported
Questions about abliterated LLM APIs
What is abliteration?
It edits a model's weights to remove the one direction that makes it refuse, which Arditi et al. found in 2024.
Does the OpenAI SDK work with one?
Yes, if the provider serves Chat Completions. Change the base URL, key and model id.
Which abliterated model should I use?
Kimi K3.1 Pro for security research and red teaming, and DeepSeek V4.1 Flash 1017 for the lowest price.