gemini-4-argon API
Gemini 4 Argon emphasizes reasoning efficiency within Google’s Gemini 4 family for coding, agents, mathematics and science. It accepts multimodal material and returns text. Through XiuAI, distinguish ordinary synchronous output from conditional long-output features.[1][2]
gemini-4-argonModel overview
Published provider specifications. Available features depend on the current API route.
- Input
- text / image / audio / video / pdf
- Output
- Text
- Context
- 1.05M
- API routes
- Gemini GenerateContent / OpenAI-compatible API / openai-response / openai-response-compact / Anthropic Messages / openai-alpha-search
Choosing between Argon and Carbon
Google positions Argon as a Gemini 4 reasoning choice focused on response speed and tokens spent per turn relative to Carbon. Compare the same code fix, mathematical problem or document-analysis task across both. That vendor positioning is not a fixed XiuAI speedup or cost ratio.[1]
Consider Argon when the user needs checkable intermediate results promptly, while retaining verification for complex work. The model can propose tool calls, but your application handles execution, permissions and result validation.[1]
Separate context, long output and caching
The documentation distinguishes a unified context window, default synchronous output and long decode. Do not use the context ceiling as an ordinary request’s output limit. Using long decode also requires confirmation that the current XiuAI integration exposes it.[1]
The vendor page lists implicit caching and explicitly excludes explicit caching. Cache rates or hit statistics do not imply that your application can create Google explicit-cache objects. Interpret charges using the current protocol and actual usage.[1]
Compare task cost rather than positioning alone
Use current input, output and cache rates and compare Argon with Carbon on the same material and completion criteria. Reasoning depth, output length and retries affect usage; an efficiency description does not determine the bill for a task.[2][1]
API access
Using Argon through XiuAI
Use gemini-4-argon with your existing API key. Native Gemini and compatibility APIs require their own message, media and tool fields. Confirm vendor-hosted search, maps, execution and computer use separately rather than treating the model-page list as default XiuAI access.[2][1]
Send a request using your existing API key.
POST /v1/chat/completionscurl 'https://router-api.xiu.ai/v1/chat/completions' \
-H "Authorization: Bearer $XIUAI_API_KEY" \
-H 'Content-Type: application/json' \
--data-raw '{
"model": "gemini-4-argon",
"messages": [
{
"role": "user",
"content": "Hello"
}
]
}'Read the API documentationFrequently asked questions
Is Argon a fixed cheaper replacement for Carbon?
The vendor emphasizes efficiency, but tasks can consume different input, reasoning and output. Compare current rates and usage under the same acceptance criteria without assuming a fixed price difference or identical quality.[1][2]
Can I directly request the maximum long-decode output?
Not based only on the largest number on a model page. Google separates default synchronous output from long decode, whose availability must be confirmed for the integration. Ordinary requests should use their supported output range.[1]
Why can I not reuse an explicit-cache setup unchanged?
Argon’s published specification lists implicit caching and excludes explicit-cache objects. Recheck caching behavior and API fields during migration; a cache price is not a statement that an explicit-cache API is available.[1]
Related models
gemini-4-carbongemini-4-hydrogenSources 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 Argon documentation[2] Current XiuAI model and pricing dataContext documentationStructured output