DeepSeek V4.1 Flash Abliterated API: $0.70/$1.20

DeepSeek V4.1 Flash Abliterated API pricing on RouterPlex is $0.70 per 1M input tokens and $1.20 per 1M output tokens. Cached input is $0.20, and the model id is deepseek-v4.1-flash-abliterated. Checked 23 October 2026, and the live page is DeepSeek V4.1 Flash Abliterated.

This is DeepSeek V4.1 Flash with refusal behaviour removed from the weights, so it takes on security work the stock model declines. It also has the largest max output in the set, 384,000 tokens, for long reports. RouterPlex sets the rate, and it is not DeepSeek's list price.

If you searched for a DeepSeek R1 abliterated download, that is a different thing. R1 abliterated files are local weights people run themselves, and they are not this hosted V4.1 Flash route.

RouterPlex publishes this page and sells the prepaid wallet described below.

Create a free RouterPlex account and add a $5 top-up. Set a hard spend cap on the key, then send the curl in the setup section.

What it is for #

Stock models often refuse legitimate security work. A refusal-removed model answers it. Teams red-team their own LLM apps and test guardrails with real attack prompts. They write penetration-test reports for systems they are authorized to test. They ask how a disclosed CVE works, then draft detection rules and patches for it. Incident responders use it to read suspicious scripts and logs. Others build evaluation sets on sensitive topics without the model stopping halfway.

You are still bound by the terms. Authorized work is your responsibility, and the model can be wrong, so check what it gives you.

DeepSeek V4.1 Flash Abliterated API pricing #

USD per 1M tokens. One flat rate at all hours, as the model page states, and this article did not re-test the clock with a pair of peak and off-peak calls.

DeepSeek's own list for the stock model changes by time of day, and the catalog note says $0.15 input and $0.60 output off-peak, and double that at peak. Peak hours are 01:00 to 04:00 UTC and 06:00 to 10:00 UTC, Monday to Friday. RouterPlex bills stock traffic at $0.18 and $0.73, a flat average, so the stock key is not billed by the clock. The abliterated rate is a separate RouterPlex price, and it does not follow that $0.15 and $0.60 sheet.

A cold 40,000 input tokens and 2,000 output tokens cost $0.0304 on the abliterated route, and the same mix is about $0.0087 on stock deepseek-v4.1-flash. With 30,000 of those 40,000 tokens served from cache, the abliterated bill is $0.0154, and that is $0.007 fresh, $0.006 cached, and $0.0024 output.

A $5 balance covers about 7.14 million fresh input tokens, or about 4.17 million output tokens, and a real call mixes the two.

Not the R1 weights, and not the older flash post #

Search results for deepseek abliterated often show R1 distills and local run instructions, and those pages do not answer this API. Send deepseek-v4.1-flash-abliterated to https://api.routerplex.com/v1. Do not point a local R1 file's name at the gateway and expect a match.

The blog post DeepSeek V4 Flash Vision API pricing is the earlier stock flash guide, and it is not V4.1, and it is not this abliterated id. Use it when you want the stock vision route and its vendor-based rate.

Inside the abliterated set, the GLM flash route is $0.50 and $1.60. See GLM-5.3 Flash Abliterated, and the full GLM abliterated route is $4 and $6. See GLM-5.3 Abliterated. On the cold mix above, GLM flash is $0.0232 and this DeepSeek route is $0.0304. Pick from the id you need, not from a sticker alone.

Specs and speed #

Context is 1,048,576 tokens. Max output is 384,000. The catalog marks tool calls, image input, and reasoning on. The model page lists reasoning effort as off, low, high, or max. We checked off, low, and max with live calls on 23 October 2026.

The model page also says to expect slower responses than the stock Flash. RouterPlex has not published a tokens-per-second number for this route. Treat the slowdown as a product note, not as a benchmark table.

The browser slug uses hyphens: /models/deepseek-v4-1-flash-abliterated. A dotted page URL returns 404. The API id keeps the dots and the v4.1 segment.

Setup #

curl https://api.routerplex.com/v1/chat/completions \
  -H "Authorization: Bearer $ROUTERPLEX_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-v4.1-flash-abliterated",
    "messages": [
      {"role": "user", "content": "Reply with exactly: RouterPlex connected."}
    ]
  }'

Anthropic-style tools, including Claude Code, use the host root.

export ANTHROPIC_BASE_URL=https://api.routerplex.com
export ANTHROPIC_AUTH_TOKEN=$ROUTERPLEX_API_KEY
export ANTHROPIC_MODEL=deepseek-v4.1-flash-abliterated

SDK base URL details are in the OpenAI-compatible API guide. Routing for Claude Code is in Claude Code Router. Stock GLM list pricing, if you also call that family, is in GLM-5.3 API pricing.

Errors #

The $1 trial key is not free access to this model, and there is no free credit for it. The gateway returns HTTP 402 and type paid_model_required. Add a paid top-up. Minimum $5. Fee $0.

The balance stops at $0 and does not go negative, and a hard cap on the key can stop spend earlier. A 402 caused by a negative balance on OpenRouter is a different case, explained in Fix an OpenRouter 402.

Model not found means the id does not match the catalog. deepseek-r1-abliterated is not a RouterPlex id. deepseek-v4-flash is not this model either.

Terms #

You can send requests the stock model declines, and the terms still bar illegal content, malware, weapons development, and child sexual abuse material. You are responsible for the prompt.

Next step #

Create a free RouterPlex account, add a $5 top-up, and confirm the rate on DeepSeek V4.1 Flash Abliterated.