Gemini 3.6 Flash has arrived. Here is a summary of what has been officially confirmed and what remains unknown.

While browsing X, a post from the official Gemini account about "Gemini 3.6 Flash" appeared in my feed.

At first, I was surprised, thinking, "It's already at 3.6?"

Whenever a new AI model is announced, performance comparisons and rumors spread quickly on social media.

However, this time, it felt like there was a mix of what could be understood from the official post and what has not yet been disclosed.

Therefore, in this article, I will organize and separate the facts confirmed by official information, what I have verified on my own screen, and what remains unknown at this time.

Confirming Gemini 3.6 Flash via the official post

On July 22, 2026, the official Google Gemini X account introduced "Gemini 3.6 Flash."

The post included a short demo video showing how to extract natural textures from photos and use them for prints or design materials.

Furthermore, the official account mentioned in a reply that "3.6 Flash" can be selected from the Gemini model selection menu.

Up to this point, these are the facts confirmed by the official post.

This article is based on that post. However, I will not conclude that it has "evolved significantly" by extending it to performance capabilities not mentioned in the post.

What we learned from this post

What caught my eye in this demo was not just answering text, but the way it reads textures contained in photos and transforms them into creative materials.

For example, it shows a workflow where you pick up textures like wood grain, fabric, or walls from photos taken in your surroundings and treat them as patterns or design elements.

The official post introduces Gemini 3.6 Flash as an example of how it quickly understands context and helps you build your own creative tools.

However, this is just one demo. It does not show that other uses, such as Japanese writing ability, coding, or long-text comprehension, have improved in the same way.

All we know is that "this is the kind of usage the official team showcased."

I also checked my own Gemini screen

After reading the official post, I opened my Gemini screen on the same day and checked the model selection menu.

What appeared on my screen were "3.1 Flash-Lite," "3.5 Flash," and "3.1 Pro." The currently selected model was 3.5 Flash, and 3.6 Flash was not yet visible.

Although the official announcement states that it is available from the model selection menu, at least on my screen, it was not yet selectable. I have not been able to confirm the scope of availability or the conditions for it to appear.

Based on this result alone, I cannot conclude that it is "unavailable in Japan" or "unavailable for paid plans." Whether it is a phased rollout or if there are differences based on region or account remains unconfirmed.

Therefore, I decided that the most accurate way to record this for now is: "The official announcement has been confirmed, but it has not yet arrived on my screen."

There is still much that is unknown

I also checked the official documentation for Google DeepMind and the Gemini API.

As of the check on July 22, 2026, the publicly available model pages and API documentation were centered on descriptions of 3.5 Flash. Within the scope of my check, I could not find detailed model cards or API descriptions for 3.6 Flash.

Therefore, I will not yet make definitive statements on the following points:

• Benchmark performance

• API availability and pricing

• Context length

• Changes in Japanese language quality

• Changes in coding performance

• Changes in image understanding and generation quality

• Target regions, target plans, and rollout timing

Filling in the gaps where information is missing with data from previous models or social media posts might make the text easier to read, but it would deviate from the facts.

For newer models, it seems necessary to leave the unknown parts as they are.

I want to keep social media information and official information separate.

This time, the first information I saw was a post on X. However, even on the same social media platform, posts from official accounts and predictions from third parties must be treated separately.

When a new model name is added, explanations like 'faster than before,' 'stronger at coding,' and 'same price' start to flood in all at once.

It is possible that those are correct. Even so, at a stage where they cannot be confirmed through official documents or actual hardware, I will not use them as facts in my articles.

This time, I checked the official X post, the Google DeepMind model page, the Gemini API model documentation, and my own model selection screen in order.

In this check, while there was an introduction on the official X account, detailed specifications and API documentation were not yet available.

When reading breaking news, it is important to look not only at what was announced but also at how much documentation is available.

What I want to actually try from here on

Once I can select 3.6 Flash on my screen, I will first try tasks close to my daily work.

• How natural the Japanese drafts are

• Whether it can organize the main points of long documents without losing their meaning

• How accurately it can read the content of images

• Whether it can reproduce the workflow of creating textures or design materials from photos

• How well it can handle simple code corrections when requested

Rather than just measuring speed, I want to find situations where I feel it is worth switching from 3.5 Flash for this specific task.

Instead of deciding which is better compared to ChatGPT, checking the suitability for each task is more useful for practical usage.

At this stage, this is not yet an evaluation article. I will record the verification after it becomes available in a separate article.

Summary at this point

Gemini 3.6 Flash was confirmed via the official Google Gemini X post. A demo showing how to extract natural textures from photos and turn them into design materials has also been released.

On the other hand, 3.6 Flash did not yet appear on my Gemini screen. Detailed specifications and API information are also not yet available in the official documents I checked this time.

When I see a new name, I immediately want to know its performance. I was the same way at first.

However, what remains after this confirmation is a more modest conclusion.

Judge only by what is known, and leave what is unknown as it is.

Once 3.6 Flash becomes available to select, I will begin verification from there.

Confirmation date: July 22, 2026

Referenced official information:

• Google Gemini official X post: https://x.com/GeminiApp/status/2079597340223320373

• Google DeepMind Gemini Flash: https://deepmind.google/models/gemini/flash/

• Gemini API model list: https://ai.google.dev/gemini-api/docs/models

*Model names, availability, and specifications are subject to change. I will reconfirm official information before publication.

Also read.