decuff GTM brain

decuff provides hosted open-weight AI models through an API for developers and users seeking fewer refusals and limited data retention.

category

Open-weight model inference API

market profile

  • text: Developers can choose among hosted open-model APIs instead of operating model infrastructure themselves.
  • text: Inference buyers compare token prices, model choice, compatibility and data retention.

personas

  • title: AI application developer
    pains:
    • Model refusals interrupt legitimate application workflows.
    • Serving open models requires operational capacity and inference infrastructure.
  • title: Technical founder
    pains:
    • Production traffic requires predictable capacity and manageable token costs.
    • Sensitive user prompts raise questions about provider retention.

icp

  • text: Developer-led software teams embedding text generation into apps or coding workflows, from solo builders to production teams.
  • text: Organizations with privacy-sensitive prompts and an appetite for hosted open-weight models.

message pillars

  • text: Access open-weight models through one endpoint.
  • text: Prompts are cleared after the response.
  • text: Start on a free allocation and pay for additional use.

objections

  • objection: Can we trust the model with sensitive prompts?
    response: Clarify that servers read prompts to generate replies and ask for the retention terms and security controls before production use.
  • objection: Will model quality and cost suit our workload?
    response: Test representative tasks on the free allocation, then compare per-token rates and model limitations.

competitors

  • name: Venice
    why: Offers private and uncensored model inference through an API.
  • name: Featherless
    why: Hosts open models for developers.
  • name: OpenRouter
    why: Provides a unified API to multiple model providers.

keywords

  • uncensored LLM API
  • open weight model API
  • zero retention AI API
  • hosted AI inference
  • pay per token LLM
  • private AI API

voice

  • direct and irreverent
  • privacy-conscious
  • developer-oriented

signals

  • text: A team is evaluating hosted open models or migrating an application away from model refusals.
  • text: A production AI application needs higher throughput or a single invoice.

stack

  • Qwen
  • DeepSeek

pricing

model: freemium
starting price: $0; paid developer plan $20/month or prepaid $10 minimum

motion

hybrid

channels

  • X
  • developer documentation
  • SEO

confidence

medium

Founders

  • name: Maceo Cardinale Kwik
    title: Founder/CEO
    role: CEO
    bio: Founder and CEO of decuff, an API for uncensored AI: abliterated open-source models served at scale with zero data retention. Previously a software engineer at Traba, where I built Datost on nights and weekends before joining YC. Datost reached $400K ARR in eight weeks. B.S. in Computer Science from Virginia Tech in two years, 4.0 GPA.

Back to YC companies