Models & pricing / gemini-4-carbon · Google
Google

gemini-4-carbon API

Gemini 4 Carbon is Google’s reasoning model for complex software engineering, mathematical and scientific problems, and multi-step agents. It analyzes multimodal material and outputs text. Evaluate long-task cost on XiuAI using its exact ID, current protocol and billing conditions.[1][2]

gemini-4-carbon

Model overview

Published provider specifications. Available features depend on the current API route.

Input
text / image / audio / video / pdf
Output
Text
Context
1M
API routes
Gemini GenerateContent / OpenAI-compatible API / openai-response / openai-response-compact / Anthropic Messages / openai-alpha-search

Making complex work checkable

Evaluate Carbon for problems involving code, reasoning, charts or several sources. Supply evidence, goals and constraints, and ask the response to retain key support and unresolved questions. Mathematical, scientific and engineering conclusions still need application or human verification.[1]

Compared with Argon, Carbon’s vendor positioning emphasizes demanding reasoning. Define correctness, turnaround and budget, then compare like-for-like samples. The vendor’s “most capable” wording is not an XiuAI guarantee for every request.[1]

Multimodal and long-output boundaries

The official input range includes text, images, audio, video and PDFs, with text output. Video understanding does not deliver generated video, and the page does not list image generation, audio generation or Live API support. Check encoding and constraints for the current API when sending media.[1]

Unified context and long decode are different concepts, and default synchronous output has its own range. Vendor long decode and hosted tools do not automatically become ordinary XiuAI request capabilities. Confirm access before setting their parameters.[1]

Long reasoning and cache costs

For long-source tasks, distinguish input, output and cache usage under consistent pricing conditions. The vendor lists implicit caching only; cache hits or lower cache rates do not establish explicit-cache creation. Reconcile charges through the current catalog and request records.[1][2]

API access

Using Carbon through XiuAI

Call gemini-4-carbon with your existing key and balance. Validate the required protocol and input before adding tools or other media. Search, maps and execution on Google’s platform require their own API support and are not gained merely by changing the model name.[1][2]

Send a request using your existing API key.

POST /v1/chat/completions
curl 'https://router-api.xiu.ai/v1/chat/completions' \
  -H "Authorization: Bearer $XIUAI_API_KEY" \
  -H 'Content-Type: application/json' \
  --data-raw '{
  "model": "gemini-4-carbon",
  "messages": [
    {
      "role": "user",
      "content": "Hello"
    }
  ]
}'
Read the API documentation

Frequently asked questions

Must I choose Carbon for an interactive application?

No. If Argon or another available model meets the task, compare equivalent quality, time and cost requirements. Carbon’s demanding-reasoning positioning is not a default answer for every application.[1][2]

Can the entire context window be used for the answer?

No. Context includes input and output, while ordinary synchronous output and long decode have separate conditions. Configure output for the request mode and current integration rather than deriving one limit from another.[1]

Does Carbon directly generate images, speech or video?

The published output modality is text. It can analyze media input, but producing images, speech or video requires a generation model or a separately integrated tool. Multimodal input is not a media-generation promise.[1]

Related models

gemini-4-argongemini-4-hydrogen

Sources and updates

Specifications reference provider documentation; prices and availability come from the current XiuAI catalog. Specifications checked: 2026-11-11

Content reviewed: 2026-11-16

[1] Google Gemini 4 Carbon documentation[2] Current XiuAI model and pricing dataContext documentationStructured output