Abliterated Model Router | AllModels
Route abliterated language models across providers with automatic fallbacks, modality-aware selection, and one place to compare availability and token prices.
Route abliterated language models across providers with automatic fallbacks, modality-aware selection, and one place to compare availability and token prices.

One endpoint for every abliterated model.
p50 — ms
— tok/s
modelDeepSeek V4.1 Flash Abliterated
Language
JavaScript · OpenAI SDK
import OpenAI from "openai";
const client = new OpenAI({
apiKey: process.env.ALLMODELS_API_KEY,
baseURL: "https://api.allmodels.io/oai",
});
const response = await client.responses.create({
model: "deepseek/deepseek-v4.1-flash-abliterated",
input: "Open the pod bay doors.",
});
console.log(response.output_text);
A fast reasoning model served by several independent providers. AllModels keeps text routes in the fallback pool and sends image requests only to a provider that can actually read them.
Loading live DeepSeek provider routes…
Open the full DeepSeek model page →Abliterated GLM 5.3 deployments from multiple providers share one public model ID. Route through the default waterfall or explicitly choose a provider when a particular price or limit matters.
Loading live GLM provider routes…
Open the full GLM model page →Kimi K3.1 Pro with its refusals removed. It reads text and images and has a 1M-token context window. AllModels adds new providers for it to the same model ID as they come online.
Loading live Kimi provider routes…
Open the full Kimi model page →More models available
DeepSeek, GLM and Kimi are only the starting point. Browse the full catalogue as new abliterated models and competitive provider routes become available.
We don’t retain your prompts or outputs, or use them to train models. Provider policies differ, so review them if retention matters to you. Read our ZDR policy
Decision gates Coming soon
Add an optional gate in front of any model. You write the question a decision model answers. Each possible answer gets its own route, whether that is a different model or a dropped request. AllModels runs the decision and the inference in a single API call, so your application calls one gated model ID.
Your rules, one model ID
Abliterated models will answer almost any prompt, so the controls belong in your routing rather than in the model. A gate checks each request before inference: you decide which prompts reach an abliterated model, which go to a standard model with its refusal behaviour intact, and which are dropped.
Gated model
model: "deepseek/deepseek-v4.1-flash-CUSTOM-ID"
Decision model
Which category does this prompt fall into?
Routes
allowed→ deepseek/deepseek-v4.1-flash-ultrafast
sensitive→ deepseek/deepseek-v4.1-flash-abliterated
prohibited→ drop · 403 Forbidden
Federated model access
AllModels continuously brings new abliterated models and competitive inference providers into one routing layer. Get the latest models, provider diversity, and capacity failover through a single security-hardened API—without rebuilding your integration each time the market moves.
Chat CompletionsPOST /oai/chat/completions
Use messages, streaming, function tools, and the familiar OpenAI chat schema.
ResponsesPOST /oai/responses
Use input items, streaming events, reasoning controls, structured output, custom tools, and hosted web search.
Automated fallbacks
AllModels routes each request across compatible providers. If one is unavailable or capacity constrained, traffic falls back automatically—without changing your model ID or application code.
ModalityImage and multimodal requests only go to providers that support them, while text requests can use every eligible text route.
CapacityWhen one provider is overloaded or unavailable, AllModels automatically tries the next eligible route for the same model.
PriceAllModels keeps provider pricing current and can route to the lowest-cost eligible option as new inference providers become available.
Texto de allmodels.io tal como lo publicó, leído de su página y mostrado aquí con el formato de El Autómata. El original, en allmodels.io · Volver a la noticia de El Autómata