---
title: "DeepSeek V4.1 Flash 1017 Defense | AI Models | Inference.net"
description: "Run DeepSeek V4.1 Flash 1017 Defense (552B, DeepSeek) on Inference.net. 1.04M context. tool-calling, structured outputs, vision input, reasoning. Requires approved access. Deploy with an OpenAI-compatible API and pay-per-token pricing."
url: "https://inference.net/models/deepseek-v4.1-flash-1017-defense/"
---

DeepSeek Reasoning Model

# DeepSeek V4.1 Flash 1017 Defense

A refusal-removed variant of the October 2026 DeepSeek V4.1 Flash checkpoint. Same 552B-parameter multimodal Mixture-of-Experts with image input, tool calling, and a 1M-token context. Available only to teams approved for privileged model access.

## Requires approved access

DeepSeek V4.1 Flash 1017 Defense has reduced safety behavior. Each team needs an approved request before it can call this model; requests without approval return HTTP 403. A person reviews every request, and the decision arrives by email.

[How privileged access works](https://inference.net/privileged-access/) [Privileged Model Access Terms](https://inference.net/model-access/deepseek-v4.1-flash-1017-defense/)

Model details

Developed by DeepSeek

Model family DeepSeek

Size 552B parameters

Base model DeepSeek V4.1 Flash 1017

Architecture Mixture-of-Experts

Context window 1016K tokens

Max output 384K tokens

Model API pricing Per 1M tokens

Input $0.30

Cache read $0.007

Output $1.00

[See all pricing](https://inference.net/pricing/)

DeepSeek V4.1 Flash 1017 Defense is the October 2026 DeepSeek V4.1 Flash checkpoint with its refusal behavior removed, one of our Defense models (also known as "abliterated" or "abilated" models). The weights come from the open mrfakename/DeepSeek-V4.1-Flash-1017-abliterated release on Hugging Face. It answers requests the base model declines, so it is built for work where conventional refusals get in the way: red-team evaluation, trust-and-safety tooling, security research, and fiction.

We serve it on the same engine configuration as the base model, so the context window, max output, input modalities, tool calling, structured outputs, and reasoning-effort levels on this page are the base model's. We have not run separate benchmarks on these weights; see the base DeepSeek V4.1 Flash 1017 page for its published results.

Access is per team and per model. Every request is reviewed by a person, the decision arrives by email, and an approval reaches the API within 45 seconds. Zero Data Retention teams can apply, and request and response content is not kept.

```
import OpenAI from "openai";

const openai = new OpenAI({
  baseURL: "https://api.inference.net/v1",
  apiKey: "<YOUR_API_KEY>",
});

const completion = await openai.chat.completions.create({
  model: "deepseek-v4.1-flash-1017-defense",
  messages: [
    {
      role: "user",
      content: "What is the meaning of life?"
    }
  ],
  stream: true,
});

for await (const chunk of completion) {
  process.stdout.write(chunk.choices[0]?.delta.content as string);
}
```

Prompt caching

1. Keep the prompt prefix stable and byte-identical across calls.
2. Do not send prompt_cache_key. These routes do not support it and may reject the request.
3. Check usage.prompt_tokens_details.cached_tokens in the response to confirm cache reads.

Use cases

### Red-team and safety evaluation

Generate adversarial and jailbreak prompts and harmful example content to evaluate AI systems you own or are authorized to test.

### Defensive security research

Write exploit code, phishing simulations, and attack tooling for use only against systems you own or have written authorization to test.

### Trust-and-safety tooling

Produce labeled abusive, hateful, or violent examples to train classifiers and build moderation tooling.

### Fiction and creative writing

Write dark or mature fiction without refusals. Sexually explicit content is allowed only when it is fictional, depicts consenting adults who are not real people, and reaches only adults, where it is lawful.

Frequently asked questions

## Related models

[See all models](https://inference.net/models/)

Model

[DeepSeek V4.1 Flash 1017 1016K input context Vision Tools Reasoning $0.30 / $1.00](https://inference.net/models/deepseek-v4.1-flash-1017/)

[DeepSeek V4.1 Flash 1016K input context Vision Tools Reasoning $0.30 / $1.00](https://inference.net/models/deepseek-v4.1-flash/)

[Kimi K3 1024K input context Vision Tools Reasoning $3.00 / $15.00 2 variants](https://inference.net/models/kimi-k3/)

[MiniMax M3.1 Flash 977K input context Vision Tools Reasoning $0.10 / $0.50](https://inference.net/models/minimax-m3.1-flash/)

## Request access to DeepSeek V4.1 Flash 1017 Defense

Approval is per team and per model, and the decision arrives by email. Questions? Email [trust@inference.net](mailto:trust@inference.net).

[Read the terms](https://inference.net/model-access/deepseek-v4.1-flash-1017-defense/) [How access works](https://inference.net/privileged-access/)