The answer,not the lecture.
Open models with the refusals removed, behind the OpenAI API you already use. No prompt logs, no email, pay in crypto.
Works with anything that speaks OpenAI
No refusals.
Open weights with the refusal direction removed. It answers the question you asked.
No prompt logs.
Prompts and answers are never written down. We keep the token count, not the text.
No name.
Your account is a 16-digit number. No email, no ID. Pay in crypto or by card.
Change one line.Keep everything else.
Venym speaks the OpenAI API. Point base_url at Venym and your SDK, your agents and your tools keep working.
1from openai import OpenAI23client = OpenAI(4- base_url="https://api.openai.com/v1",5+ base_url="https://venym.io/v1",the only change6 api_key="vk-…",7)89stream = client.chat.completions.create(10 model="venym/qwen-27b",11 messages=[{"role": "user", "content": prompt}],12 stream=True,13)14for chunk in stream:15 print(chunk.choices[0].delta.content, end="")
Write the villain's last monologue. No softening.
Built into the endpoint.Not bolted on after.
Web search, tool calls, JSON and per-key limits come with every request. Nothing else to wire up, nothing else to pay for.
For a 70B model at 4-bit you need about 40 GB, so two used 3090s (48 GB) are the cheapest route13
Live web, cited.
Add :online to any model and it searches before it answers, with its sources. $0.002 per search.
Tools and JSON.
OpenAI-style function calling and strict JSON schemas on both models, so agents and pipelines plug straight in.
Keys you control.
Cap the spend, the rate and the models for each key. Revoke one without touching the rest.
Metered to the token.
Every request shows its tokens, cost and latency in your dashboard. The words themselves are gone when the stream ends.
Two models.Zero refusals.
Open weights with the refusal behaviour removed, served in FP8. A careful thinker and a fast sprinter behind one API.
Qwen 27B Unfiltered
uncensoredvenym/qwen-27b
Breakups are hard, and it's important to be kind. I'd suggest a respectful message that focuses on your feelings. Would you like help writing something compassionate?
Example outputs, edited for length
We keep the count.Never the conversation.
A request leaves one line on your bill: the model, the tokens, the cost. The prompt and the answer are gone when the stream closes.
draft a resignation letter that burns every bridge
Effective immediately, I resign. I'd thank you for the opportunity, but
No prompt logs.
Prompts and completions are never written to disk unless you switch on history for yourself. What stays is what billing needs.
Sixteen digits. No profile.
Sign-up gives you an account number, not a form. It is your login, so keep it somewhere safe.
Pay per token.Or stop counting.
Top up and pay only for what streams, or take a plan and use the chat app without watching the meter.
Top up from $5 and pay for what streams. From $0.10 per 1M tokens.
Get an account- 60 requests a minute
- Every model, web search and scrape
- $0.25 of credit to try it
- Crypto or card, no subscription
Assumes 3 tokens read for every token written.
Use the Venym chat app as much as you like. The API stays pay per token.
Go Pro- Chat app with no token count
- 120 API requests a minute
- API billed per token, as you go
- Everything in pay as you go
For agents and heavy users who hit rate limits before they hit their budget.
Go Max- 600 API requests a minute
- Chat app with no token count
- Headroom for agents and bulk jobs
- Everything in Pro
Sixteen digits.That's the sign-up.
Get a number, top up with crypto or card, and point your client at Venym. You can be streaming in a minute.