NEWKimi K3.1 Pro Abliterated (40% OFF!)
Powerful cyber-capable models for security teams, served via API with zero data retention.
Hosted behind an OpenAI- and Anthropic-compatible API. You pay per token, and we keep no copy of your prompts.
Set the base URL to https://api.refuseless.com/v1 and pick a model id. The OpenAI and Anthropic SDKs, and agents built on them, work without other changes.
We remove the refusal direction from the model itself. Nothing is added to your prompt, so a new session does not bring the refusals back. How abliteration works
Prompts and outputs are not saved and not used for training. A partner provider, if one serves the request, gets it without your IP, email, or key. Privacy policy
Each score is the base model’s own published number, not a Refuseless eval. Open a model for its specs and a request to copy.
| Model | Published scores | Context | Speed | Input | Output | Details |
|---|---|---|---|---|---|---|
| GLM 5.3 AbliteratedHard coding, agent, and security work |
| 1M | 75 tok/s | $1.00 | $2.00 | |
| GLM 5.3 Flash AbliteratedEveryday and high-volume work |
| 1M | 80 tok/s | $0.12 | $0.50 | |
| DeepSeek V4.1 Flash AbliteratedCoding agents and terminal work |
| 1M | 80 tok/s | $0.12 | $0.45 |
Prices are in USD per million tokens. Score sources: Z AI's GLM-5.3 model card, Z AI's GLM-5.3-Flash release post, DeepSeek's V4.1-Flash model card.
The honest version, so you know what you are buying.
Prompts a stock chatbot declines get a normal completion. That is the one thing abliteration changes.
Abliteration removes a brake. It adds no knowledge and raises no benchmark. The model can still be wrong or vague.
The model will discuss the dangerous case. Whether your use is allowed is on you and on our terms: no attacks on systems you do not own or are not authorized to test.
Setup takes about a minute. Pick your agent:
~/.config/opencode/opencode.json%USERPROFILE%\.config\opencode\opencode.json{
"$schema": "https://opencode.ai/config.json",
"model": "refuseless/glm-5.3-abliterated",
"provider": { "refuseless": {
"npm": "@ai-sdk/openai-compatible",
"name": "Refuseless",
"options": {
"baseURL": "https://api.refuseless.com/v1"
},
"models": {
"glm-5.3-abliterated": {
"name": "GLM 5.3 Abliterated",
"tools": true,
"reasoning": true,
"limit": {
"context": 1048576,
"output": 131072
},
"variants": {
"low": {
"reasoningEffort": "low"
},
"high": {
"reasoningEffort": "high"
},
"max": {
"reasoningEffort": "max"
},
"medium": {
"disabled": true
}
}
},
"glm-5.3-flash-abliterated": {
"name": "GLM 5.3 Flash Abliterated",
"tools": true,
"reasoning": true,
"limit": {
"context": 1048576,
"output": 131072
},
"variants": {
"low": {
"reasoningEffort": "low"
},
"high": {
"reasoningEffort": "high"
},
"max": {
"reasoningEffort": "max"
},
"medium": {
"disabled": true
}
}
},
"deepseek-v4.1-flash-abliterated": {
"name": "DeepSeek V4.1 Flash 1017 Abliterated",
"tools": true,
"reasoning": true,
"limit": {
"context": 1048576,
"output": 131072
},
"variants": {
"off": {
"reasoningEffort": "off"
},
"low": {
"reasoningEffort": "low"
},
"high": {
"reasoningEffort": "high"
},
"max": {
"reasoningEffort": "max"
},
"medium": {
"disabled": true
}
}
},
"deepseek-v4.1-flash-fast": {
"name": "DeepSeek V4.1 Flash 1017",
"tools": true,
"reasoning": true,
"limit": {
"context": 1048576,
"output": 262144
},
"variants": {
"off": {
"reasoningEffort": "off"
},
"low": {
"reasoningEffort": "low"
},
"high": {
"reasoningEffort": "high"
},
"max": {
"reasoningEffort": "max"
},
"medium": {
"disabled": true
}
}
},
"minimax-m3.1-flash": {
"name": "Minimax M3.1 Flash",
"tools": true,
"reasoning": true,
"limit": {
"context": 1048576,
"output": 512000
},
"variants": {
"low": {
"reasoningEffort": "low"
},
"high": {
"reasoningEffort": "high"
},
"max": {
"reasoningEffort": "max"
},
"medium": {
"disabled": true
}
}
},
"kimi-k3.1-pro": {
"name": "Kimi K3.1 Pro",
"tools": true,
"reasoning": true,
"limit": {
"context": 1048576,
"output": 131072
},
"variants": {
"low": {
"reasoningEffort": "low"
},
"high": {
"reasoningEffort": "high"
},
"max": {
"reasoningEffort": "max"
},
"medium": {
"disabled": true
}
}
},
"deepseek-v4.1-flash-abliterated-fast": {
"name": "DeepSeek V4.1 Flash 1017 Fast Abliterated",
"tools": true,
"reasoning": true,
"limit": {
"context": 1048576,
"output": 131072
},
"variants": {
"off": {
"reasoningEffort": "off"
},
"low": {
"reasoningEffort": "low"
},
"high": {
"reasoningEffort": "high"
},
"max": {
"reasoningEffort": "max"
},
"medium": {
"disabled": true
}
}
},
"kimi-k3.1-pro-abliterated": {
"name": "Kimi K3.1 Pro Abliterated",
"tools": true,
"reasoning": true,
"limit": {
"context": 1048576,
"output": 131072
},
"variants": {
"low": {
"reasoningEffort": "low"
},
"high": {
"reasoningEffort": "high"
},
"max": {
"reasoningEffort": "max"
},
"medium": {
"disabled": true
}
}
}
}
} }
}New file: copy the whole file. Already have one: copy only the highlighted block and paste it inside the existing "provider" object, keeping your other providers.
Start opencode, type /connect, choose Refuseless and paste your key. If Refuseless isn't listed, pick Other and enter refuseless as the provider id.
Use the key from your API keys page. It starts with abl_.
OpenCode saves this key separately, and it takes priority over any apiKey in the config. If you create a new key later, run /connect again, or OpenCode keeps sending the old one and you get Invalid API key.
opencode --model refuseless/glm-5.3-abliteratedOr use /models in OpenCode and pick Refuseless › GLM 5.3 Abliterated. Send Say ok. to check it works. All 8 lineup models are in the config, so you can switch any time.
Press ctrl+t to cycle through the reasoning efforts the model offers. Until you pick one, the model's default applies.
Three volume tiers for products built on Refuseless, and higher concurrency, TPM and RPM limits on request.
Starter
+5% creditsFrom $1,000 / month
Growth
+10% creditsFrom $5,000 / month
Scale
+15% creditsFrom $10,000 / month
Extra credits on your monthly spend, usable on every model.
Reseller program detailsAn open-weight model with the direction that triggers refusals removed from its weights. It answers prompts a stock chatbot would refuse, and you call it like any other model. It still knows only what the base model knew, and it can still be wrong.
Sometimes. Abliteration lowers refusals, it does not promise zero. Our terms also forbid some uses, such as attacking systems you do not own or are not allowed to test. You decide what your own application accepts.
GLM 5.3 Abliterated for hard or long tasks, GLM 5.3 Flash Abliterated for everyday work at a lower price, DeepSeek V4.1 Flash Abliterated for coding agents. Each model page lists its scores and price.
Yes. Point an OpenAI SDK at https://api.refuseless.com/v1, or an Anthropic SDK at the same base with /v1/messages, and set the model id. Send the API key as a Bearer token. Coding agents such as OpenCode and Pi work as they are.
Not your prompts or outputs. They are processed and discarded, and never used for training. We keep account, login, and billing records. Some requests may be served by a partner provider, which receives the request without your IP address, email, or key. See the privacy policy.
No. You add credit and pay per token. New accounts sometimes get a small bonus on their first deposit, shown at signup when it is running.
By card (Visa, Mastercard) or crypto (BTC, ETH, USDT, USDC, XMR). You prepay a balance and spend it as you go.
Sign up with email or Google, add credit, and call your first model.