Models
Gemini 4 Argon: What We Know About Google’s Model
Gemini 4 Argon is Google’s stable reasoning model for complex coding, mathematics, and multi-step tasks. Google lists a 1,048,576-token context window and API pricing of $2 per million input tokens and $10 per million output tokens through November 30, 2026. From December 1, the published prices are $4 and $20, respectively.
What is Gemini 4 Argon?
Gemini 4 Argon is a Google model with low, medium, and high reasoning modes. Google documentation calls it stable; its API model ID is `gemini-4-argon`. Listed use cases include software development, scientific and mathematical work, and multi-step agent workflows.
The 1,048,576-token context window covers input and output combined. A standard synchronous response is limited to 65,536 output tokens. For longer generation, Google documents a separate long-decode mode of up to 1,048,576 tokens including input.
Supported inputs and tools
An API feature is not necessarily available in every third-party client. The client, intermediary, and specific model must all support the required tool-call format.
| Capability | According to Google documentation |
|---|---|
| Input | Text, images, video, audio, and PDF |
| Output | Text |
| Tools | Function calling, Search, URL context, code execution, File Search, and Google Maps grounding |
| Reasoning | Low, medium, and high levels |
| Not supported | Audio generation, image generation, and Live API |
Gemini 4 Argon API pricing
Google bills thinking tokens as output. Your actual cost therefore depends on reasoning volume, context, and the selected mode, not only on the visible answer. Google’s current pricing page describes caching, batch processing, and priority processing terms.
| Period | Input per 1M tokens | Output per 1M tokens |
|---|---|---|
| Through November 30, 2026 | $2 | $10 |
| From December 1, 2026 | $4 | $20 |
Is the model available in Any LLM?
Check the current Any LLM catalog. If Gemini 4 Argon is listed for your account, connect with the same key and a compatible API client. The catalog shows currently available models and pricing; the official Gemini API is Google’s direct route to the model.
Before moving a project, check supported parameters too: Google says temperature, top_p, and top_k are ignored for Gemini 4 Argon, and explicit API caching is not supported.
Who should consider Gemini 4 Argon?
Consider it for tasks that need long context, complex code, multi-step analysis, or several input types. For short, frequent requests, compare the total cost with lighter models. The most reliable approach is to test several of your own tasks and measure the full cost of each answer.