Work through a problem
Ask for an explanation, compare approaches, and refine the answer in conversation.
OPEN CHAT →Built in India · Chat + API
WRITE · CODE · BUILD
Think it through.Build it out.With BrokenGPT.
Work through ideas, draft your next piece, or untangle code in chat. Build AI into your own app with an OpenAI-compatible API. A privately routed mixture of open-source models powers both.
Account creation is free; AI usage is paid. Choose prepaid credits or Pro. Compare pricing.
1M TOKEN CONTEXT◆ANSWERS AS THEY HAPPEN◆CHAT + API, ONE BALANCE◆USAGE YOU CAN ACTUALLY SEE◆
01
Make it useful
From a first idea to your next integration.
Start with a conversation, then bring the same service into your own workflow through the API.
Built for everyday work
CHAT + APIOne account. Shared credits.Build with the API
Send your first chat-completions request with the stable broken-one alias.
READ THE DOCS →Keep track of usage
Chat and API share a balance, with usage visible in your dashboard.
COMPARE PRICING →Start with prepaid credits, or choose Pro for a monthly token allowance. Chat and API draw from one balance, with visible usage. Prepaid credits never expire.
02
For builders
Bring BrokenGPT into your app.
Use familiar chat-completions semantics, bearer API keys, server-sent events, and explicit quota headers. The docs define the supported subset.
THE CONTRACT
Change the base URL. Keep your message structure.
The public model alias stays stable while the private open-source mixture evolves behind it. That keeps client integration simple and avoids coupling applications to a specific checkpoint.
- Bearer-key auth with revocable hashed API keys.
- JSON or SSE using the same completion endpoint.
- Visible limits through request IDs and rate headers.
curl "$BROKENGPT_BASE_URL/v1/chat/completions" \
-H "Authorization: Bearer $BROKENGPT_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "broken-one",
"messages": [
{ "role": "user", "content": "Explain the tradeoff plainly." }
],
"stream": true
}'AUTHBEARER
STREAMSSE
FORMATJSON
03
Transparent by construction
No gateway censorship. Direct by design.
Prompts are not stopped by a keyword filter. Routing exists to choose the best available model tier, never to decide which questions may reach inference.
One product identity
A stable public alias sits in front of a private mixture of open-source models.
See how it worksLimits in numbers
Monthly tokens and throughput are specific per plan. Generation continues until the model finishes or reaches its context boundary.
Compare plansNo prompt gate
Every prompt reaches the configured mixture. The router selects capability, not permission.
Read the system contract04
The build sequence
Three phases. No secret leap between them.
We are starting with proven open-weight models, then publishing the work required to make them less restrictive without trading away capability, before training a BrokenGPT-native model.
PHASE 01 LIVE NOW
JUL / 2026
Open-weight foundation
Route a private mixture of open-source models through one metered chat and API path. Establish the product, usage data, and evaluation baseline before changing model behavior.
CURRENT STAGEPHASE 02TARGET
OCT / 2026
High-fidelity uncensoring
Reduce unnecessary refusals in selected open models while preserving the highest practical accuracy. Where infrastructure permits, evaluation and serving will use full-precision, non-quantized checkpoints.
RESEARCH + EVALSPHASE 03TARGET
EARLY / 2027
Our own uncensored model
Develop a BrokenGPT-native model for maximum response freedom, with its identity, training decisions, evaluations, limitations, and legal safety perimeter published alongside it.
MODEL DEVELOPMENTTHE WORK WILL BE PUBLIC
Methods, evaluations, failures, tradeoffs, infrastructure, and major model decisions will be documented in the build journal.Read the blog05
Pay as you go
Pay for the tokens you use, nothing else.
Add prepaid credits as needed, or get 10 million included tokens each paid month with Pro for $29.
01 / USAGE
Pay as you go
$0.50/ 1M input tokens
Add credits, spend them per token. No subscription, no monthly reset.
- Output rates from $2.00 / 1M tokens
- Model-specific rates published below
- Top up $5–$500 — credits never expire
- Chat + OpenAI-compatible API on one balance
- $1 in free credit when you sign up
Most popular
02 / PRO
Pro
$29/ month
10 million included tokens for individual power users, renewed each paid month.
- 10M tokens across chat + API
- 60 requests per minute
- Prepaid credits cover any extra usage
- Unused included tokens expire each month
03 / TEAMS
Enterprise
$20/ seat / month
Seat-based plan for teams that want higher throughput and priority.
- Higher rate limits + priority routing
- Team seats, managed in the billing portal
- Token usage billed to your shared credit balance
- Volume pricing on request
Pro: $29 / month with 10M tokens · prepaid credits never expire.
See the full breakdown06
Straight answers
Before you wire it into anything.
01Does BrokenGPT censor prompts?
No gateway keyword filter blocks prompts. Every request is sent to the private open-source model mixture, with routing used only to select the best inference tier.
02What powers BrokenGPT?
BrokenGPT routes requests through a private mixture of open-source models behind one stable product identity.
03Can I use an OpenAI SDK?
The chat-completions endpoint follows an OpenAI-compatible request shape for messages, sampling controls, optional caller output controls, stop sequences, and streaming. Check the API docs for the exact supported surface before treating it as a complete drop-in replacement.
04How does billing work?
Choose model-specific prepaid usage or Pro at $29/month with 10 million included tokens. Pro tokens expire at the end of each paid month; prepaid credits never expire. Both prompt and completion tokens count.
B/GPT
BRING YOUR NEXT IDEA.
Put BrokenGPT to work.
Create an account, choose prepaid credits or Pro, and start a conversation or make your first API request.
Create an account