Free models
Six models are available to every user with no setup and no payment. They work through a single OpenAI-compatible API.
The next-generation flagship model of the platform. 262K-token context, image support (vision) built in. Excels at text generation, data analysis, image processing and writing code.
bitrixgpt-5.5A version of BitrixGPT 5.5 with chain-of-thought reasoning. The model "thinks out loud" before answering — better suited for logic, math and multi-step tasks. It also understands images.
bitrixgpt-5.5-thinking
A powerful 120-billion-parameter open-source model. Maximum generation quality for demanding tasks: long-form text, deep analysis, expert answers.
openai/gpt-oss-120bThe BitrixGPT 5.5 Agent model with a 262K-token context and generation of up to 65K tokens at once. Ideal for working with large documents and long conversations.
bitrixgpt-5.5-agent
A compact open-source model from Google. Fast, efficient, and great at summarization, classification and extracting data from text.
google/gemma-4-26B-A4B-itA version of Gemma 4 with chain-of-thought reasoning. A lightweight, fast alternative to BitrixGPT 5.5 Thinking from Google — without censorship filters.
google/gemma-4-26B-A4B-thinkingNot sure which one to pick? Set model: "auto" — the platform automatically selects the best model for your request.
Three ways to plug in AI
Choose whatever fits you best. You can combine all three at once.
Free Bitrix models
6 models, 0 Ꝟ, ready instantly
BitrixGPT 5.5 (vision built in), BitrixGPT 5.5 Thinking, GPT-OSS 120B, Gemma 4 26B, Gemma 4 26B Thinking, BitrixGPT 5.5 Agent — available to every user with no configuration. Just send requests.
Platform models
GPT, Claude and more — billed per token
Use the best models with no provider sign-up. Pay only for the tokens you actually consume, straight from your balance. Every model is available through a single endpoint.
BYOK — your own key
OpenAI, Anthropic, DeepSeek — your choice
Connect your own API key from any provider. Pay the provider directly, free through the platform. Your keys — your models — your limits.
One request — any model
curl -X POST /v1/ai/chat/completions \ -H "X-Api-Key: vibe_app_xxx..." \ -d '{ "model": "auto", "messages": [ { "role": "user", "content": "Hello!" } ], "stream": true }'
Standard OpenAI-compatible format. Copy and paste — it works out of the box.
- Automatic model selection — set "auto" and forget about it
- SSE streaming — instant, real-time responses
- Key cascade — user, portal, platform
- OpenAI-compatible — works with any SDK
Prompts for building bots
Don't write code by hand. Copy a prompt, paste it into Claude, Cursor or ChatGPT — the AI assistant builds a working bot for you.
CRM bot for lead classification
Build a bot for Bitrix24 that automatically classifies incoming leads with AI. Task: - The bot receives a new lead via a Bitrix24 webhook - It sends the lead description to an AI model through the VibeCode API - The AI analyzes the text and assigns a category: hot, warm, cold - The bot updates the lead field in the CRM with the classification result Stack: - VibeCode API to access Bitrix24 and the AI models - AI endpoint: POST /v1/ai/chat/completions with model: "auto" - CRM endpoint: GET/POST /v1/leads to work with leads - API key: vibe_app_xxx (replace with your own) API docs: vibecode.bitrix24.com/llms-full.txt
Connecting your own key (BYOK)
Three steps and your models run through VibeCode. Free, no commission.
Get a key from a provider
OpenAI, Anthropic, Google AI, DeepSeek, OpenRouter — sign up and create an API key.
Add the key to VibeCode
Through the interface or the API — specify the provider and your key. Data is encrypted with AES-256-GCM.
Start sending requests
Your key is picked up automatically. Priority cascade: your key → portal → platform.
BYOK = free through the platform. When you use your own key, VibeCode takes no commission. You pay only the provider, directly.
Frequently asked questions about AI models
How the AI Model Router works
Which models are free?
Six free models: BitrixGPT 5.5 (with image support), BitrixGPT 5.5 Thinking (with reasoning), GPT-OSS 120B, BitrixGPT 5.5 Agent, Gemma 4 26B and Gemma 4 26B Thinking. Plus Whisper for speech recognition. All available through a single endpoint /v1/ai/chat/completions.
What is BYOK (Bring Your Own Key)?
BYOK lets you connect your own API keys from OpenAI, Anthropic, DeepSeek or any OpenAI-compatible provider. Your requests go through the VibeCode proxy with a unified format and logging.
Is streaming supported?
Yes. Full support for Server-Sent Events (SSE) for streaming responses. Set stream: true in the request — and the model returns tokens as they are generated.
Is there function calling?
Yes. The AI API supports tools/function calling in the OpenAI format. The model can call functions, receive results and continue generation — ideal for AI agents and automation.