The GPT-5.6 API
that costs nothing
Chat, vision, web search, tool calling and file input — fully OpenAI-compatible, streamed over SSE. Create a key and you're running in 30 seconds.
curl https://gpt.abhibots.com/v1/chat/completions \
-H "Authorization: Bearer sk-gptb-..." \
-H "Content-Type: application/json" \
-d '{
"model": "gpt-5.6",
"messages": [
{"role": "user", "content": "Hello!"}
]
}'Everything in the box
Verified live — every feature below is tested, not listed hopefully.
GPT-5.6 “Luna”
The current frontier model, self-served. 24 model aliases route to the latest — names cost nothing extra.
Web search built-in
Ask current-events questions and get live answers with real source URLs. No plugin, no extra cost.
Real streaming
Token-by-token SSE, OpenAI wire-format. Drop-in for any SDK or proxy that speaks the protocol.
Vision & OCR
Send images — screenshots, scans, photos. Reads text out of pixels, powered by real GPT vision.
Tools & JSON mode
Function calling with proper arguments, structured output, conversation state — all verified working.
File input
Attach txt, md, csv, code, text-PDFs. Extracted faithfully and inlined — no RAG loss.
Endpoints
Everything an OpenAI client expects, on this domain.
One backend, 24 names
Pick any — all free, all routed to the current best model.
Vision & files
Standard OpenAI content parts, nothing custom to learn.
# image (png / jpg / gif / webp)
{""type": "image_url", "image_url": {"url": "data:image/png;base64,..."}}
# file — text-extracted, faithful
{""type": "file", "file": {"filename": "notes.txt",
"file_data": "data:text/plain;base64,..."}}
Honest limits
| Max per message | ~120k chars (~30k tokens) |
| Larger inputs | auto-chunked server-side, recent context prioritised |
| Rate limit | generous — rotating residential egress |
| Audio in / image-gen | not available |
| Price | ₹0 — free forever |