Gemini 4 Carbon
Neural Network
Gemini 4 Carbon is Google's frontier model for programming, agentic tasks, and complex reasoning.
Specifications
- Input
- text, images, documents
- Output
- text
- Context size
- 1 048 576 tokens
- Max answer length
- 943 718 tokens
- Reasoning effort
- low, medium, high
- Capabilities
- reasoning, function calling, Batch API
Prompt cost
(per 1M tokens)
Answer cost
(per 1M tokens)
Image prompt
(per 1K tokens)
import OpenAI from 'openai';
const openai = new OpenAI({
apiKey: '<your bothub access token>',
baseURL: 'https://openai.bothub.chat/v1'
});
// Sync - Text generation
async function main() {
const chatCompletion = await openai.chat.completions.create({
messages: [{ role: 'user', content: 'Say this is a test' }],
model: 'gemini-4-carbon',
});
}
// Async - Text generation
async function main() {
const stream = await openai.chat.completions.create({
messages: [{ role: 'user', content: 'Say this is a test' }],
model: 'gemini-4-carbon',
stream: true
});
for await (const chunk of stream) {
const part: string | null = chunk.choices[0].delta?.content ?? null;
}
}
main();
How it works Gemini 4 Carbon?
Gemini 4 Carbon is Google's fourth-generation frontier model in the Gemini line, optimized for programming, agentic scenarios, and complex reasoning. It accepts text, images, audio, video, and files as input and supports long outputs, making it capable of tasks requiring the retention of large amounts of material without losing logic. In BotHub, the model is available without a VPN or foreign card: access from Russia, payment with Russian cards in rubles, and pay-as-you-go billing—tokens do not expire. Other models are available in the same window, allowing for real-time comparison and switching without rewriting integrations, while a unified OpenAI-compatible API is available for multimodal systems. Teams have access to contracts, invoices, electronic document management (EDO), and an admin panel with limits. Typical scenarios: refactoring and debugging large projects, analyzing mixed materials like screenshots, meeting recordings, and PDFs, long research reasoning chains, agentic pipelines, and automating routine engineering tasks.Frequently asked questions about Gemini 4 Carbon
You can use generated results for commercial purposes. You own all rights to the content you create. The only restriction: make sure your prompt does not include copyrighted third-party material. You are responsible for respecting the rights to any input data.
This is a text model for long and rich contexts: it accepts up to 1,048,576 tokens, works with documents and images, can reason before answering, and call functions. It is suitable for analyzing large volumes of documentation, analytics, coding tasks, and multi-step calculations.
In a single response, the model can output up to 943,718 tokens—a very large volume comparable to several books. The input window reaches 1,048,576 tokens, so entire sets of documents and related correspondence fit into a single session.
Yes, reasoning before answering is part of the model's capabilities. It can break down a task step-by-step before formulating a conclusion—this helps in analytics, mathematics, and code debugging, where the process of solving is as important as the final line.
Yes, function calling and structured JSON output are supported. The model can be connected to external tools and services, and responses can be received in a machine-readable format—this is convenient for integrations and automation where the result goes directly into another system.
According to our data, images and documents are accepted as input: you can upload a file or image and ask questions about the content. Audio and video input are not noted in the model description, so this should be verified in the documentation before building a scenario around it.
There are no separate quality assessments for the Russian language in the model's data, so we cannot promise specifics. The model is designed for working with large texts, and the level of language proficiency is best checked with your own examples—instructions, documents, and typical queries.