Introduction

On November 12, 2026, a new model Gemini 4 Hydrogen (Model ID: gemini-4-hydrogen) was generally available (GA) on the Gemini API. It's a model that emphasizes speed and efficiency in everyday coding and agent workflows, with the slogan "Cutting-edge intelligence for everyone."

The key points of this update are the following three:

• Free tier available, paid tier is $1.50 per input / $7.50 per output (per 1 million tokens)
• Added fourth thinking level xhigh
• Default model for Gemini Managed Agents' Antigravity agents changed to gemini-4-hydrogen

Especially the third point, if you are using agents without explicitly specifying the model, there is a possibility that behavior, pricing, and response tendencies may change, so it's necessary to confirm.

📌 Who is impacted

• Developers using Gemini 3.8 Flash (need to check API changes)
• Users of Antigravity agents in Gemini Managed Agents without specifying the model
• Users adjusting thinking levels for coding support and agents
• Individual developers and learners wanting to try the new model on the free tier

Overview of Changes

Change Details

1. Gemini 4 Hydrogen is GA

Item	Content
Model ID	gemini-4-hydrogen
Status	General Availability (GA)
Free Tier	Available for use
Paid Tier (Input)	$1.50 / 1 million tokens
Paid Tier (Output)	$7.50 / 1 million tokens
Intended Use Cases	Daily coding, agent workflows

Pricing details can be confirmed on the official pricing page (/docs/pricing#gemini-4-hydrogen).

2. Default model change for new thinking level xhigh

A fourth level, xhigh, has been added to the traditional three levels.

Level	Notes
low	Same as before
medium	Same as before
high	Default (no change)
xhigh	Newly added

The default remains high, so existing code without any specifications will not change its behavior to xhigh. For details, refer to /docs/thinking#understanding-thinking-levels.

3. Default model change for Antigravity agents

The Antigravity agent in Gemini Managed Agents (/docs/antigravity-agent) has changed the default model to gemini-4-hydrogen.

⚠️ Breaking Change (similar behavior change)
If you are using the Antigravity agent without specifying a model, it will now operate with a different model than before, potentially changing its behavior, costs, and response tendencies. Since the official documentation does not specify the extent of these changes, it is recommended to test your workload to confirm.

4. Migration Guide

The API changes and migration steps from Gemini 3.8 Flash are summarized in "What's new in Gemini 4 Hydrogen" (/docs/whats-new-gemini-4-hydrogen). If you are using Gemini 3.8 Flash, please make sure to check the API changes.

Note: Correction of Gemini Decision notation (minor impact)

In the November 11th changelog entry, "Gemini Decision" was corrected to "Gemini Decision 1". The model ID gemini-decision-1, the linked content, and the content of the Decisions API (which targets text, images, audio, video, and PDF, and performs fast, calibrated binary/category/rubric judgments with zero output tokens) have not changed. No action is required.

Impact and Response

Actions for Developers

1. Gemini 3.8 Flash Users: Read the migration guide and identify the API changes.
2. Antigravity Agent Users: If the model is not specified, you may be affected by the default change. Verify the quality, cost, and response tendency, and specify the model if you want to fix the behavior.
3. Cost Estimation: Recalculate using $1.50 per input and $7.50 per output for 1M tokens.
4. Review of thinking level: Try xhigh for difficult tasks. However, the default remains high.
5. Update Internal Documents: Update the model list, pricing, and thinking settings (including the default model for xhigh and Antigravity).

💡 Tips
Since there is a free tier, it is low risk to first run existing prompts with gemini-4-hydrogen and compare the output differences with the current model.

Example of Cost Estimation

This is an estimate for processing 20 million input tokens and 4 million output tokens per month.

• Input: 20 × $1.50 = $30
• Output: 4 × $7.50 = $30
• Total: Approximately $60 per month

※ For details such as billing for thinking tokens, please check the official pricing page.

Code Example

The following is an example of explicitly specifying the thinking level and model. Please note that parameter names and syntax may differ depending on the SDK version, so always check the official documentation (this is a reference example showing the structure only, as the input data does not include the code itself).

Before（State without considering the model and thinking level）

After（Explicitly specify model and thinking level）

Summary

• Gemini 4 Hydrogen（gemini-4-hydrogen）is GA. It's a fast and efficient model for coding/agent use, with a free tier available, paid at $1.50 per input / $7.50 per output (per 1M tokens).
• Added xhigh thinking level. The default remains high, so there's no impact on existing code.
• Default model for Antigravity agent has changed. If not specified, test changes in behavior, cost, and response tendencies.
• Gemini 3.8 Flash users should check the migration guide for API changes.
• Corrections to Gemini Decision 1's notation are only documentational and require no action.

First, try existing workloads with the free tier, and it's recommended to start by checking whether Antigravity is used and the status of model specifications.