Gemini 4 Argon lands on AI Gateway: 1M-token context, thinking levels replace thinking budgets

Google's Gemini 4 Argon can now be called through AI Gateway, aimed at coding, agentic work, and complex reasoning. The model was not previously among the models available on AI Gateway, so callers needed another access path. The AI Gateway page gives the model name google/gemini-4-argon and lists the AI SDK, Chat Completions, Responses, and Anthropic Messages APIs as ways to call it.

The model replaces thinking budgets with low, medium, and high thinking levels, so callers pick a level rather than a number. This kind of control previously appeared as a thinking budget that required a specific value. The thinking level is set with the reasoning parameter in the AI SDK and reasoning_effort in Chat Completions, with other controls described under reasoning.

The model reads dense diagrams, spreadsheets, and layout-heavy documents, suiting tasks such as a financial analysis spread across several spreadsheet tabs. Such tasks previously required breaking layout information into text or structured fields first. The model accepts text, image, PDF, and video input and supports tool calling, and on coding tasks it carries longer tasks through multiple steps.

The report covers Gemini 4 Argon's availability on AI Gateway, its input types, thinking-level control, and calling methods. It suits engineers choosing a model for coding agents or long-context multimodal tasks. The material states no follow-up experiment or deployment recommendation.

The material states only that the model targets coding, agentic work, and complex reasoning, and lists input types and calling methods; it gives no benchmark scores, comparison models, or cost figures. The mapping between thinking levels and thinking budgets is also not explained.