We use cookies to enhance your experience, provide personalised content, and analyse our traffic; see our Cookie Policy.
google/gemini-4-carbonGoogle's top reasoning model, and most of its work is invisible
Gemini 4 Carbon is Google's top reasoning tier, released on 10 November 2026. Google built it for complex software engineering, maths and science, and multi-step agents. It takes text, images, audio, video and PDFs, returns text, and reads up to 1,048,576 tokens of context.
Last updated Nov 19, 2026
It always thinks before it answers, and that thinking cannot be switched off. The thinking is billed as output tokens and you do not see it. In our own test, an exact counting puzzle came back right after 7,905 billed output tokens, of which about 7,723 were thinking: 97.7% of what was billed never appeared on screen, for a visible answer of 182 tokens. So the output price per million tokens tells you little about what a request will cost. Test it on your own prompts before you plan a budget.
On Tchavi you pay with mobile money, no bank card and no KYC. One account and one API key open 100+ models, and your credits never expire.
It thinks through hard problems before answering, and you cannot turn that off
A codebase or a stack of long documents in a single call
Mixed input in one request, text out
Function calling and structured output for your agents
Gemini 4 Carbon always reasons before it answers, and that reasoning is billed as output tokens even though it is not shown. In our test, 7,723 of 7,905 billed output tokens (97.7%) were thinking, for an answer of 182 tokens. The more demanding the problem, the more it thinks.
No. Reasoning is always on for this model. If you want short, fast replies, pick a lighter model from the catalogue.
Up to 1,048,576 tokens in. Output is capped at 32,768 tokens per request on Tchavi, thinking included in what is billed, so leave headroom for it.
Text, images, audio, video and PDFs, and it returns text.
Swap the base URL for Tchavi's, use your Tchavi key and set model: "gemini-4-carbon" on /v1/chat/completions.
Yes. Top up your Tchavi account with Orange Money, MTN MoMo, M-Pesa or another mobile money operator. No bank card is needed. Credits land as soon as the payment is confirmed and are deducted as you use them.
Yes. Keep your current OpenAI SDK, just change the base URL to Tchavi's (shown in the API tab) and use your Tchavi key. The rest of your code stays the same.
import Tchavi from '@tchavi/sdk';
const client = new Tchavi({ apiKey: 'API_KEY' });
const response = await client.chat.completions.create({
model: 'gemini-4-carbon',
messages: [
{
"role": "user",
"content": "Hello!"
}
],
});
console.log(response.choices[0].message.content);
console.log('Credits used:', response.tchavi.credits_used);Every parameter this model accepts on the API. Check the SDK for full typings.
modelRequiredID of the chat model (e.g. `gemini-4-carbon`).
messagesRequiredConversation messages. Each item has a `role` (system/user/assistant) and a `content` that's either a plain string or an array of content blocks (`text`, `image_url`, `document_url`).
temperature1Sampling temperature between 0 and 2. Higher values = more random.
max_tokensMaximum number of tokens to generate in the response.
max_completion_tokensOpenAI alias for `max_tokens`. Either field works; if both are sent, `max_completion_tokens` wins.
streamfalseIf true, partial message deltas are streamed as Server-Sent Events.
toolsFunction-calling toolbox. Pass an array of `{ type: "function", function: { name, description, parameters } }`. Returned tool calls land in `choices[].message.tool_calls`. Only models with `supportsTools: true` accept this.
tool_choiceHow the model picks among tools: `"auto"` (default), `"none"` to force a plain text reply, or `{ type: "function", function: { name } }` to force a specific tool. Only with `supportsTools: true`.
response_formatConstrain the output. Pass `{ type: "json_object" }` to force valid JSON (the model still needs a JSON instruction in your prompt). Only models with `supportsJsonMode: true` honour this.
Estimated over 30 days, for a typical message: ~500 input + ~150 output tokens, billed to the thousandth of a credit.