Skip to main content

DeepSeek V4.1 Flash Abliterated

DeepSeek V4.1 Flash with refusals removed, served by Lutheos with zero data retention and a 1M-token context; if no supplier route is available it falls back to our own DGX Spark.

DeepSeek V4.1 Flash Abliterated

1M context with tool use, reasoning and image input.

  • Input, per 1M tokens€0.12
  • Cached input, per 1M tokens€0.016
  • Output, per 1M tokens€0.47
6,200 tokens, priced at the published rateThis request€0.001164
All models and prices
Create a key

Specification

Context window
1M
Max output
131.1K
Capabilities
vision, tools, reasoning
Family
DeepSeek
Model id
deepseek-v4.1-flash-abliterated

First request

Request
curl https://api.zdr.sh/v1/chat/completions \
  -H "Authorization: Bearer $ZDR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"deepseek-v4.1-flash-abliterated","messages":[{"role":"user","content":"Hello"}]}'
ResponseSample response
data: {"choices":[{"delta":{"content":"Hello"}}]}
data: {"choices":[{"delta":{"content":"!"}}]}
data: {"usage":{"prompt_tokens":12,"completion_tokens":2},
       "cost":"0.00000238","retention":"zero"}
data: [DONE]

A sample response. The token counts and the cost are the sample's, priced at this model's rates.

OpenAI-compatible https://api.zdr.sh/v1Anthropic-compatible https://api.zdr.sh

Keep the key in the ZDR_API_KEY environment variable, never in a file a tool can commit.

Pricing and routing

Charged per token at this rate, whichever serving tier answers the request.

The rateEach bucket the ticket prints is billed per token at the rate it shows.
Per token
Cached inputThe cached rate applies when the provider serves the prefix from its cache, and a cache write adds nothing.
Its own rate
The routeThe route that answers a request never changes its price.
Same price
Require zero retentionA model id with the :zdr suffix keeps a request off rented capacity, and the request is refused rather than served by one.
:zdr suffix

Policy

This is an abliterated variant: the refusal behaviour the provider trained into it is removed. It runs under the same retention rules and the same acceptable use policy as every other model.