Seedance 2.5 — 15% below official pricing, 57% cheaper than Higgsfield and Fal. Draft mode is live · Wan 3.0 — 35% below official list price.

Guides, updates and engineering notes from ArgoLink on building with one API for every AI model. Blog

Googlegemini-4-carbon

Google's most capable Gemini reasoning model, built for complex software engineering, math and science

$0.6/ 1M input tokens-70%$2official price

  • 1M context
  • Up to 64K output
  • Effort low to high
  • Always thinks
  • Reads images
Modality
Text
Endpoint
/v1/chat/completions
Model ID
gemini-4-carbon
Provider
Google

Gemini 4 Carbon API overview

Documented endpoint

Requests go to /v1/chat/completions, /v1/responses or /v1/messages, and the gateway converts each to Google's Gemini API. Google's own format, /v1beta/models/gemini-4-carbon:generateContent, works too.

Model provider

Gemini 4 Carbon is published by Google and is called through the API model ID gemini-4-carbon.

Output and limits

1,048,576-token context; up to 65,536 output tokens; text and images in, text out.

Gemini 4 Carbon API pricing

Gemini 4 Carbon is billed 70% below the official list price on every metered dimension below.

ItemArgoLink rateOfficial priceYou save
Input tokens$0.6 / 1M tokens$2-70%
Cached input tokens$0.03 / 1M tokens$0.1-70%
Output tokens$3 / 1M tokens$10-70%

Reasoning tokens are billed as output tokens.

Create an API key

Gemini 4 Carbon API reference

Quick start

Create an API key in the console, then use the Base URL, endpoint, and request body shown on this page.

Base URL

  • https://argolink.io/v1For OpenAI SDKs and OpenAI-compatible clients.
  • https://argolink.ioFor Anthropic SDKs, Claude Code and other Anthropic Messages clients. Leave out /v1: the client adds /v1/messages itself.
  • https://argolink.ioFor Google's Gen AI SDKs and other Gemini API clients. Leave out /v1beta: the client adds it itself.

Endpoints

  • POSTRecommended/v1/chat/completionsConverted to Google's Gemini API.
  • POSTResponses/v1/responsesConverted to Google's Gemini API.
  • POSTAnthropic Messages/v1/messagesConverted to Google's Gemini API.
  • POSTGemini API/v1beta/models/gemini-4-carbon:generateContentGoogle's own format, for the Gemini SDKs; authenticate with x-goog-api-key.

Authentication

Send your API key as a Bearer token in the Authorization header.

POST /v1/chat/completions
curl -sS https://argolink.io/v1/chat/completions \
  -H "Authorization: Bearer $ARGOLINK_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gemini-4-carbon",
    "messages": [{"role": "user", "content": "Explain what this model is good at in one sentence."}],
    "stream": false
  }'

Protocols

Chat Completions

Recommended: converted to Google's Gemini API

OpenAI SDKs and OpenAI-compatible clients work as they are: the gateway converts each request to Google's Gemini API.

POST /v1/chat/completions
curl https://argolink.io/v1/chat/completions \
  -H "Authorization: Bearer $ARGOLINK_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gemini-4-carbon",
    "messages": [{"role": "user", "content": "Compare optimistic and pessimistic locking in three bullet points."}],
    "reasoning_effort": "high"
  }'
With an imagePOST /v1/chat/completions
curl https://argolink.io/v1/chat/completions \
  -H "Authorization: Bearer $ARGOLINK_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gemini-4-carbon",
    "messages": [{
      "role": "user",
      "content": [
        {"type": "text", "text": "What does this chart show?"},
        {"type": "image_url", "image_url": {"url": "data:image/png;base64,iVBORw0KGgo..."}}
      ]
    }]
  }'
FieldValue
messagessystem, user, assistant and tool turns; a user turn may carry image_url parts, each a data URL
thinkingalways on; leave it out
reasoning_effortlow · medium · high
max_completion_tokensup to 65,536
streamtrue streams server-sent events
service_tiernot supported; leave it out

max_completion_tokens counts thinking tokens too.

Responses and Anthropic Messages

Converted to Google's Gemini API

Codex, Claude Code and other clients that send these formats reach the same model: the gateway converts each request to Google's Gemini API.

ResponsesPOST /v1/responses
curl https://argolink.io/v1/responses \
  -H "Authorization: Bearer $ARGOLINK_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gemini-4-carbon",
    "input": "Compare optimistic and pessimistic locking in three bullet points."
  }'
Anthropic MessagesPOST /v1/messages
curl https://argolink.io/v1/messages \
  -H "x-api-key: $ARGOLINK_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gemini-4-carbon",
    "max_tokens": 4000,
    "messages": [{"role": "user", "content": "Compare optimistic and pessimistic locking in three bullet points."}]
  }'
FieldValue
inputtext and input_image parts; an image must be a data URL, because image links are dropped
messagestext, tool and base64 image blocks; image URLs are dropped
reasoning.effortthe same values as reasoning_effort
output_config.effortthe same values; sent as medium when left out

Gemini 4 Carbon full request parameters

ParameterTypeRequirementWhat it does
modelstringrequiredThe catalog identifier for this page: gemini-4-carbon.
messagesarrayrequiredSystem, user, assistant and tool turns. User content may carry image_url parts, each a data URL.
thinkingobjectoptionalThinking: always on; leave it out.
reasoning_effortstringoptionalReasoning depth: low · medium · high.
max_completion_tokensintegeroptionalCaps the answer: up to 65,536. max_tokens works too.
tools · tool_choicearray · string or objectoptionalFunction tools and how the model picks one.
streambooleanoptionaltrue streams chunks whose choices[].delta carry the text.
service_tierstringnot supportedThere is no faster tier for this model; leave it out.

Create an API key

Gemini 4 Carbon API compared with similar models

ModelModel IDProviderRateBest for
Gemini 4 Carbongemini-4-carbonGoogle$0.6 / 1M input tokensGoogle's most capable Gemini reasoning model, built for complex software engineering, math and science
GPT-6 Astragpt-6-astraOpenAI$2.3 / 1M input tokensOpenAI's most capable model, built for the hardest end-to-end work
GPT-5.6 Terragpt-5.6-terraOpenAI$0.46 / 1M input tokensThe GPT-5.6 model that balances intelligence and cost
GPT-5.6 Solgpt-5.6-solOpenAI$0.92 / 1M input tokensThe GPT-5.6 flagship for complex professional work

More models from this provider

Other models in this modality

Gemini 4 Carbon API FAQ

How do I call Gemini 4 Carbon?

Create an API key in the console and POST https://argolink.io/v1/chat/completions with model set to gemini-4-carbon and your key in the Authorization header as a Bearer token. /v1/responses and /v1/messages work too; the gateway converts every request to Google's Gemini API.

How much does Gemini 4 Carbon cost?

$0.6 per million input tokens and $3 per million output tokens, 70% below the official rates. Cached input costs $0.03 per million. Reasoning tokens are billed as output tokens.

What are Gemini 4 Carbon's context and output limits?

1,048,576-token context; up to 65,536 output tokens; text and images in, text out.

How do I control Gemini 4 Carbon's reasoning?

It always thinks. Set reasoning_effort to low, medium or high.

Can Gemini 4 Carbon read images?

Yes: send them as image_url parts, each a data URL, on /v1/chat/completions. Image links are dropped on every format, so send each image inside the request.

Create an API key

Related guides

Welcome Builders

Who runs ArgoLink, refunds and how your data is handled

Request Refund

Submit a request and review its status in Top-up Orders.