[Dagon Inference](https://inference.dagon.so/) / Anthropic Messages API compatible endpoint for uncensored models

# An Anthropic Messages API compatible endpoint for uncensored models

Published 2026-11-11 · Last updated 2026-11-11

Yes. Dagon Inference serves uncensored (abliterated) open-weight models on an Anthropic Messages API compatible endpoint, `POST https://inference.dagon.so/v1/messages`: set your Anthropic SDK's base URL to `https://inference.dagon.so` (no `/v1`), pass a Dagon Inference key as `x-api-key` and use a model id such as GLM-5.3 Abliterated (`glm-5.3-abliterated`, served on Messages by Silk Compute, from $2.20 input and $4.40 output per 1M tokens). Streaming and client tools work as in the Anthropic API. The models are open-weight models with refusal behavior removed, not Claude, and you pay per token in USDC, with no sign-up and no subscription.

## What are the base URL, endpoint and headers?

- Base URL for Anthropic SDKs: `https://inference.dagon.so`. The SDK adds `/v1/messages`.
- Endpoints: `POST https://inference.dagon.so/v1/messages`, and `POST https://inference.dagon.so/v1/messages/count_tokens` to count tokens for free.
- Key: `x-api-key: gk_...`, as the Anthropic SDKs send it; `Authorization: Bearer gk_...` works too, and wins if both are sent.
- `anthropic-version` is forwarded as you send it.
- Errors on this endpoint use the Anthropic error shape.

The official Anthropic SDKs also read `ANTHROPIC_BASE_URL` and `ANTHROPIC_API_KEY` from the environment, so code that builds its client without arguments switches by setting those two variables and the model id.

Same plug, new socket

The gateway answers on four paths: /v1/chat/completions and /v1/responses for OpenAI SDKs, /v1/messages for Anthropic SDKs, and /v1/messages/count\_tokens, a free pass-through. Your client code stays as it is: point its base URL at the gateway and give it a gateway key.

## Python (anthropic)

```
# pip install anthropic
import os
import anthropic

client = anthropic.Anthropic(base_url="https://inference.dagon.so", api_key=os.environ["DAGON_KEY"])
msg = client.messages.create(
    model="glm-5.3-abliterated",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Explain this stack trace."}],
)
print("".join(block.text for block in msg.content if block.type == "text"))
```

## TypeScript (@anthropic-ai/sdk)

```
// npm install @anthropic-ai/sdk
import Anthropic from "@anthropic-ai/sdk";

const client = new Anthropic({ baseURL: "https://inference.dagon.so", apiKey: process.env.DAGON_KEY });
const msg = await client.messages.create({
  model: "glm-5.3-abliterated",
  max_tokens: 1024,
  messages: [{ role: "user", content: "Explain this stack trace." }],
});
console.log(msg.content.filter((b) => b.type === "text").map((b) => b.text).join(""));
```

## curl

```
curl https://inference.dagon.so/v1/messages \
  -H "x-api-key: $DAGON_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "content-type: application/json" \
  -d '{"model": "glm-5.3-abliterated", "max_tokens": 1024, "messages": [{"role": "user", "content": "Hello"}]}'
```

Add `"stream": true` for server-sent events in the Messages event format.

## Which models answer on /v1/messages?

A Messages request goes only to the providers that serve its model on Messages, so the providers and the price can differ from the model's cheapest route on Chat Completions. The abliterated models are marked removed; models marked standard or reduced keep some or most of their refusals.

| Model | Model id | Refusals | Input | Output | Providers on Messages |
| --- | --- | --- | --- | --- | --- |
| [GLM-5.3 Abliterated](https://inference.dagon.so/models/glm-5.3-abliterated) | `glm-5.3-abliterated` | removed | $2.20 | $4.40 | Silk Compute |
| [DeepSeek V4.1 Flash Abliterated](https://inference.dagon.so/models/deepseek-v4.1-flash-abliterated) | `deepseek-v4.1-flash-abliterated` | removed | $0.11 | $0.44 | Lutheos and Silk Compute |
| [GLM-5.3 Flash Abliterated](https://inference.dagon.so/models/glm-5.3-flash-abliterated) | `glm-5.3-flash-abliterated` | removed | $0.264 | $1.10 | Silk Compute |
| [Abliterated Large Flash](https://inference.dagon.so/models/abliterated-model-large-flash) | `abliterated-model-large-flash` | removed | $1.10 | $2.20 | abliteration.ai |
| [Abliterated Model](https://inference.dagon.so/models/abliterated-model) | `abliterated-model` | removed | $1.10 | $3.30 | abliteration.ai |
| [Abliterated Large v2](https://inference.dagon.so/models/abliterated-model-large-v2) | `abliterated-model-large-v2` | removed | $3.30 | $5.50 | abliteration.ai |
| [Kimi K3.1 Pro Abliterated](https://inference.dagon.so/models/kimi-k3.1-pro-abliterated) | `kimi-k3.1-pro-abliterated` | removed | $5.50 | $16.50 | Silk Compute |
| [DeepSeek V4.1 Flash](https://inference.dagon.so/models/deepseek-v4.1-flash) | `deepseek-v4.1-flash` | standard | $0.022 | $0.176 | Silk Compute |
| [GLM-5.3 Flash](https://inference.dagon.so/models/glm-5.3-flash) | `glm-5.3-flash` | standard | $0.088 | $0.462 | Lutheos and Cheaper Inference |
| [Kimi K3.1 Pro](https://inference.dagon.so/models/kimi-k3.1-pro) | `kimi-k3.1-pro` | standard | $0.55 | $5.50 | Silk Compute |
| [Hy4](https://inference.dagon.so/models/hy4) | `hy4` | standard | $1.76 | $3.52 | Silk Compute |

Prices are per 1M tokens in USD at the cheapest provider serving each model on Messages, with the 10% markup included, at price version 2026-11-11.4. A request that falls back to another provider is billed at that provider's rate plus the markup; its receipt names the provider. Live: [/v1/models](https://inference.dagon.so/v1/models) (each offer lists its endpoints).

## What does the endpoint support, and what does it refuse?

- Fields: `model`, `messages`, `max_tokens`, `system`, `stop_sequences`, `stream`, `temperature`, `top_p`, `top_k`, `tools`, `tool_choice`, `thinking` and `metadata`. `max_tokens` is required; any other non-null field is `400 unsupported_parameter`.
- Client tools, with `tool_choice` `auto`, `any`, `none` or a named tool. Server tools (web search, web fetch, code execution, MCP) are refused.
- Images as `image` blocks with a `base64` source, on models that take image input.
- No `anthropic-beta` value is accepted: a request with one is `400 unsupported_parameter`, so a client that requires a beta does not work.
- Token counting is free and served by abliteration.ai, so it works for the models abliteration.ai serves on Messages.

Full reference: [Anthropic Messages docs](https://inference.dagon.so/docs/messages).

## How do you get a key and pay?

Create a key with `curl -X POST https://inference.dagon.so/v1/keys` (no email or password) or in the [console](https://inference.dagon.so/app/), and fund it with USDC from $5. An agent can also pay per request with [x402](https://inference.dagon.so/x402-llm-api) on `/v1/messages`, without a key. Every model's price is on [the pricing page](https://inference.dagon.so/abliterated-llm-pricing).

We do not monitor or log your content, and the acceptable use policy applies: use it lawfully and only against systems and people you own or have permission to test. We suspend keys that break it.

## Common questions

### Is there an Anthropic Messages API compatible endpoint for uncensored models?

Yes. Dagon Inference answers `POST https://inference.dagon.so/v1/messages` in the Anthropic Messages format for 7 abliterated models: GLM-5.3 Abliterated (`glm-5.3-abliterated`, served on Messages by Silk Compute, from $2.20 input and $4.40 output per 1M tokens), DeepSeek V4.1 Flash Abliterated (`deepseek-v4.1-flash-abliterated`, served on Messages by Lutheos and Silk Compute, from $0.11 input and $0.44 output per 1M tokens), GLM-5.3 Flash Abliterated (`glm-5.3-flash-abliterated`, served on Messages by Silk Compute, from $0.264 input and $1.10 output per 1M tokens), Abliterated Large Flash (`abliterated-model-large-flash`, served on Messages by abliteration.ai, from $1.10 input and $2.20 output per 1M tokens), Abliterated Model (`abliterated-model`, served on Messages by abliteration.ai, from $1.10 input and $3.30 output per 1M tokens), Abliterated Large v2 (`abliterated-model-large-v2`, served on Messages by abliteration.ai, from $3.30 input and $5.50 output per 1M tokens) and Kimi K3.1 Pro Abliterated (`kimi-k3.1-pro-abliterated`, served on Messages by Silk Compute, from $5.50 input and $16.50 output per 1M tokens). Point an Anthropic SDK at `https://inference.dagon.so` with a Dagon Inference key and pay per token in USDC.

### Does it serve Claude?

No. It serves open-weight models from abliteration.ai, Refuseless, Experiential Labs, Lutheos, Silk Compute and Cheaper Inference; the Messages endpoint shares only Anthropic's request and response format. Dagon Inference is not affiliated with Anthropic.

### Can I keep my Anthropic SDK code?

Yes, if it uses the supported fields: change the base URL, the key and the model id. Requests with an `anthropic-beta` header, server tools or file sources are refused with `400 unsupported_parameter`.

### Which header carries the key?

`x-api-key`, as the Anthropic SDKs send it, or `Authorization: Bearer`. If both are sent, `Authorization` wins.

## Create a key in a minute.

[Get a key](https://inference.dagon.so/app/)

## Related guides

- [Uncensored OpenAI-compatible API for security research, refusals removed](https://inference.dagon.so/cybersecurity-llm-api)
- [Smart contract audit with an uncensored LLM](https://inference.dagon.so/smart-contract-audit-llm)
- [No-KYC LLM API: use it without KYC and pay with crypto (USDC)](https://inference.dagon.so/pay-for-llm-api-with-usdc)
- [Use an uncensored LLM with the OpenAI or Anthropic SDK](https://inference.dagon.so/uncensored-llm-openai-sdk)
- [Abliterated LLM API: models, providers and pricing](https://inference.dagon.so/abliterated-llm-pricing)
- [Providers that host abliterated (refusal-removed) open-weight models through an API](https://inference.dagon.so/abliterated-model-api-providers)
- [Prepaid LLM API key for AI agents, paid in USDC](https://inference.dagon.so/buy-llm-credits-from-an-agent-wallet)
- [x402 LLM API: AI agents pay per request in USDC on Solana](https://inference.dagon.so/x402-llm-api)