Providers that host abliterated (refusal-removed) open-weight models through an API

Which abliterated models does each provider host?

US dollars per 1M tokens through Dagon Inference: the provider's price plus the 10% markup, at price version 2026-11-11.4. A provider's own direct prices and terms may differ. Live: /v1/models.

What is an abliterated model?

Abliteration edits an open-weight model's weights to remove the direction that drives refusals, so it needs a model whose weights are available: a closed model such as GPT or Claude cannot be abliterated by anyone outside its lab. The models above are labelled from our live probes, which show their refusal behavior as removed; individual requests may still be declined or answered poorly, and we have not benchmarked their quality. abliteration.ai's own abliterated build of GLM-5.3 is abliterated-model-large-v2; it is post-trained by abliteration.ai, so it can behave differently from glm-5.3-abliterated.

How do you reach every provider with one key?

Create a key with curl -X POST https://inference.dagon.so/v1/keys (no email or password), fund it with USDC from $5, and call any model id above at https://inference.dagon.so/v1 with an OpenAI SDK, or at https://inference.dagon.so with an Anthropic SDK (see the Anthropic Messages guide). The gateway picks the cheapest healthy provider serving the model on the endpoint you call; each model page lists every provider's price, and the pricing page has the whole table.

What do the providers say they keep?

abliteration.ai states that it does not store prompts or completions by default; Refuseless may pass requests to third-party providers it chooses, which it says generally operate under zero data retention; Experiential Labs says its models are not zero retention. Lutheos says its GPU partner keeps no prompts or replies; Silk Compute says it does not store prompts or outputs and serves its models without content filters. Cheaper Inference keeps request content unless the request asks for zero data retention; ours always do, and it says it then keeps none and sends the request only to providers that keep none. We cannot verify or control any of that.

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