Abliterated Model Router now live! Try it today free.

Published by AllModels

OpenRouter abliterated models

OpenRouter has no abliterated models. Here is where to call abliterated versions of the models it serves, with the same SDK, model IDs and OpenRouter-style routing for uptime, speed and price.

Cover image: a silver railway points switch thrown to a new track, beside the title OpenRouter Abliterated Models
Keep the OpenRouter-style model IDs and provider options, and switch to abliterated versions of the same models.

Does OpenRouter have abliterated models?

No. As of November 2026, OpenRouter has no abliterated models. An abliterated model is an open-weight LLM whose refusal behavior has been removed by editing its weights, so it answers requests the original model would refuse. See the strongest abliterated models for how they compare.

OpenRouter serves the original GLM 5.3, GLM 5.3 Flash, DeepSeek V4.1 Flash, Kimi K3.1 Pro and Qwen 3.8 27B, but not their abliterated versions.

Abliterated versions of current models typically need hundreds of gigabytes of GPU memory, so they can't usually be run locally without heavy quantization. For high-quality inference, you need an API.

Abliterated versions of OpenRouter's models

AllModels serves an abliterated version of each of these models. Its model IDs follow OpenRouter's, with -abliterated added.

OpenRouter model IDs and their abliterated versions on AllModels

Original model IDs from OpenRouter's model list; abliterated IDs from the AllModels catalogue, checked November 2026.
ModelOpenRouter (original)AllModels (abliterated)
GLM 5.3z-ai/glm-5.3z-ai/glm-5.3-abliterated
GLM 5.3 Flashz-ai/glm-5.3-flashz-ai/glm-5.3-flash-abliterated
DeepSeek V4.1 Flashdeepseek/deepseek-v4.1-flash-1017deepseek/deepseek-v4.1-flash-abliterated
Kimi K3.1 Promoonshotai/kimi-k3.1-promoonshotai/kimi-k3.1-pro-abliterated
Qwen 3.8 27Bqwen/qwen3.8-27bqwen/qwen3.8-27b-abliterated

Most of these models are served by several specialist providers, whose prices and speeds differ widely. The table below is live from the AllModels router.

DeepSeek V4.1 Flash Abliterated by provider on AllModels

Live from the AllModels router. Provider prices per million tokens; speed is median output tokens per second over recent production requests.
ProviderZDRInput / 1MOutput / 1MSpeed
RefuselessYes$0.13$0.47122 tok/s
Experiential LabsNo$0.24$0.95228 tok/s
LyberNo$0.31$1.04663 tok/s
HopperYes$0.32$1.27243 tok/s
AbliterationYes$1.06$2.11251 tok/s
SolCard InferenceYes$1.32$2.6488 tok/s
audnNo$3.17$5.2862 tok/s

OpenRouter-style routing for uptime, speed and price

Like OpenRouter, AllModels is a router. Most abliterated models are served by several providers, and AllModels chooses one for each request, judging every provider on its recent production requests.

  • Uptime. Failover is on by default. If a provider cannot be reached, times out, rate-limits the request or returns a server error, AllModels tries the next provider for the same model before output starts. Providers that keep failing are moved to the back of the order.
  • Speed. Add :nitro to the model ID to send the request first to the healthy provider with the fastest typical response.
  • Price. By default, abliterated models go first to the healthy provider with the lowest typical price. :floor or sort: "price" asks for this explicitly, and max_price sets a ceiling.
  • Control. Use order, only and ignore to choose providers, and allow_fallbacks: false to turn failover off.

Moving from OpenRouter

AllModels follows OpenRouter's conventions, so moving an OpenRouter call to an abliterated model takes two changes: set the OpenAI SDK base URL to https://api.allmodels.io/oai, and add -abliterated to the model ID.

  • The same provider object. order, only, ignore, allow_fallbacks, sort: "price", max_price and zdr work as they do on OpenRouter. Other fields are accepted and ignored.
  • The same shortcuts. :nitro and :floor work on every model ID.
  • ZDR. Send "provider": { "zdr": true } to use only providers with a zero data retention policy.

See the provider routing docs for every option.

Comparison

OpenRouter and AllModels for abliterated models

Both are OpenAI-compatible routers with the same provider options. Only AllModels serves abliterated models.

Models and features as documented on each company's own website, docs and API. Checked November 2026.
ProductAbliterated modelsOpenAI SDKProvider routingPricing
AllModelsYesGLM 5.3, GLM 5.3 Flash, DeepSeek V4.1 Flash, Kimi K3.1 Pro, Qwen 3.8 27BYesYesprovider object, :nitro and :floorPay as you go
OpenRouterNoYesYesprovider object, :nitro and :floorPay as you go

OpenRouter in brief

OpenRouter

The best-known general LLM router. It serves the original GLM 5.3, DeepSeek V4.1 Flash, Kimi K3.1 Pro and Qwen 3.8 27B, but no abliterated models.

Pick it if: you want the original models, or models beyond abliterated LLMs.

Call an abliterated model the way you call OpenRouter

The same OpenAI SDK, a different base URL and an abliterated model ID. Pick a client, an API and a language; highlighted lines are the AllModels base URL, the model and the provider preference.

Client
Mode
Language

OpenAI SDK, Chat Completions JavaScript · openai

Set the OpenAI SDK base URL to https://api.allmodels.io/oai and call an abliterated model by its AllModels ID. Read the LLM quickstart.

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.ALLMODELS_API_KEY,
  baseURL: "https://api.allmodels.io/oai",
});

const completion = await client.chat.completions.create({
  model: "deepseek/deepseek-v4.1-flash-abliterated",
  messages: [{ role: "user", content: "Explain abliteration in one sentence." }],
});

console.log(completion.choices[0].message.content);

Compatible interfaces

Keep the SDK you already use

https://api.allmodels.io/oai/chat/completions

OpenAI Chat Completions

Point the official OpenAI SDK at https://api.allmodels.io/oai and keep calling chat.completions.create, now with abliterated model IDs. Responses come back as JSON or stream as server-sent events.

Read the LLM quickstart

https://api.allmodels.io/oai/responses

OpenAI Responses

Use responses.create with the same client and key. The Responses API is stateless, so send the full conversation with each request.

Read the LLM quickstart

Quick answers

OpenRouter abliterated model questions

Does OpenRouter have abliterated models?

No. As of November 2026, no model in OpenRouter's model list is abliterated. AllModels serves abliterated GLM 5.3, GLM 5.3 Flash, DeepSeek V4.1 Flash, Kimi K3.1 Pro and Qwen 3.8 27B.

Is there an OpenRouter abliterated models API?

Not from OpenRouter. AllModels is an OpenAI-compatible API for abliterated models that uses OpenRouter-style model IDs and the same provider options. Set the base URL to https://api.allmodels.io/oai and call a model such as deepseek/deepseek-v4.1-flash-abliterated.

What is the abliterated version of an OpenRouter model ID?

On AllModels, add -abliterated to the ID: z-ai/glm-5.3 becomes z-ai/glm-5.3-abliterated. DeepSeek V4.1 Flash is deepseek/deepseek-v4.1-flash-abliterated.

Can I use OpenRouter's provider routing with abliterated models?

Yes. AllModels reads OpenRouter's provider object: order, only, ignore, allow_fallbacks, sort, max_price and zdr. The :nitro and :floor suffixes work too. See the provider routing docs.

Move an OpenRouter call in three steps
  1. 1
    Keep the OpenAI SDK. Use your AllModels API key in place of your OpenRouter key.
  2. 2
    Change the base URL. Point the SDK at https://api.allmodels.io/oai.
  3. 3
    Add -abliterated to the model ID. For example deepseek/deepseek-v4.1-flash-abliterated. Keep your provider options as they are.

Build with AllModels

Abliterated models, the OpenRouter way.

Start with free credits and call abliterated models from several providers with the OpenAI SDK.

AllModels is an independent service and is not affiliated with, endorsed by, or sponsored by the providers whose models it routes to.