Gemini 3.8 Flash Explained: It Is In the Gemini App, But Only If You Pay

Google shipped its third Flash model in six weeks on September 2, 2026, and this one behaves differently from the last. That day Google introduced Gemini 3.8 Flash and 3.8 Flash Cyber, calling 3.8 "our best reasoning and coding model yet, at the same speed and low cost of 3.7". Three weeks earlier, Gemini 3.7 Flash did not reach the Gemini app at all. This one does. It just does not reach the free plan.

Key Takeaways

• Google names exactly two consumer plans. Its announcement says 3.8 Flash "is available to Google AI Pro and Ultra subscribers across the Gemini app, AI Mode in Google Search and Gemini in Google Sheets".
• The free plan is not named, and on October 9, 2026 it was moved DOWN. The Gemini subscriptions page carried "Access to 3.6 Flash" under the $0 tier on September 2, September 19 and as late as October 9 at 07:14 GMT, then "Access to 3.5 Flash-Lite", with 3.6 Flash moved to the $4.99 AI Plus tier. Read live on October 19, 2026 the cards name no versions; 3.7 and 3.8 appear on no consumer plan row either way.
• Google AI Plus at $4.99 a month is not named either. The announcement's consumer line skips it.
• There is still no release note for it, and still no way to pick it. Google has published three Gemini release notes since -- 2026.09.10, 2026.09.30 and 2026.10.13 -- and the string "3.8" appears in none of them, read October 13, 2026. Its model list now lives in a different help article, About Gemini models, which names the app's models Flash-Lite, Flash, Pro and Argon and carries no version numbers either.
• This is broader than 3.7 Flash, not narrower. 3.7 reached consumers only inside Tasks in Gemini, the agent then called Gemini Spark. 3.8 reaches the chat app itself, for paying subscribers.
• The improvements are coding, agent and professional-domain work. Google's numbers cover software engineering, finance and legal agent benchmarks, and a 54.9% score on HLE-Verified. It publishes nothing about better conversation.
• Google warns the model uses more tokens. It says 3.8 Flash "works harder", executing extra reasoning steps and calling tools repeatedly, and tells developers who want efficiency to stay on 3.7 Flash.
• The second model, 3.8 Flash Cyber, is not for you. It is restricted to vetted security defenders through Google's new Fairwind Program.

Do You Get Gemini 3.8 Flash? Availability by Plan

If you pay for Google AI Pro or Ultra, yes, in three places. If you are on the free plan or on Google AI Plus, Google does not say you do. Here is every surface where you might look for it, and what Google actually published on September 2, 2026:

One thing that table cannot tell you is whether the free plan is excluded on purpose or simply not mentioned yet. Google enumerated its consumer audience in a sentence with three products and two plan names in it, which is a strong signal, but it is not the same as writing "not available on the free plan". The plan page is the surface that would confirm it, and on October 9, 2026 that page answered the question in the harshest way available: it moved the $0 tier down to 3.5 Flash-Lite.

What Google's Other Pages Still Say

Four Google surfaces that should describe this model do not mention it. That is worth knowing before you go hunting for a setting that does not exist. Checked on September 2, 2026, the day of the announcement:

This is now a pattern rather than an oversight. Google has stopped putting Flash point versions in its consumer help pages: the help article calls the middle model "Gemini 3 Flash" and says outright that "model names, versions, and availability may change". So there may be no version number for you to see anywhere in the product, and the announcement may be the only place the string "3.8 Flash" ever appears for a consumer.

The practical consequence: do not go looking for a 3.8 Flash option in your model menu. If you pay for Pro or Ultra, the change is made for you, and if you do not, there is nothing to find.

How This Compares To the Last Two Flash Releases

Three Flash models in six weeks, three different answers to "do I get it". Putting them side by side is the fastest way to see where you stand:

Read left to right, Google's Flash releases have moved off the free plan and stayed off it, and on October 9, 2026 the free plan lost the one it had. The version Google's plan page now sells to a free user is 3.5 Flash-Lite, from before any of these three.

What Gemini 3.8 Flash Is Actually Better At

Code, agents and professional analysis. Not conversation. Google's announcement is organised around benchmarks that measure software work and specialist domains, and it does not claim the model is a better chat partner:

There is one line in the announcement that a normal person can feel, and it is a downside. Google writes that 3.8 Flash "works harder": on complex tasks it "exhibits greater diligence, executing extra reasoning steps, and calling tools iteratively", and "at times, the model might use more tokens to maximize performance". For a developer paying per token, that is a cost note. For you, sitting in front of a chat box, more reasoning steps means waiting longer for an answer to a question that did not need them. Google's own advice for anyone who cares more about speed than depth is to stay on 3.7 Flash, which it says "remains fully supported for efficiency-first workloads".

The Price, and the Price After January

Consumers pay nothing extra, because 3.8 Flash arrives inside plans you already have. Developers pay $0.75 per million input tokens and $3.75 per million output tokens, the same introductory rate 3.7 Flash launched at.

The number worth recording is in the announcement's footnote, because it is a doubling with a date on it: "Introductory price expires on December 31, 2026. Starting January 1, 2027, $1.50/1M input tokens and $7.50/1M output tokens will apply." If you use a tool built on the Gemini API, that is the date its running cost changes, and the announcement itself is the only place Google publishes it.

What Gemini 3.8 Flash Cyber Is

A security model that finds and fixes software vulnerabilities, locked to vetted defenders. You cannot get it, and you would have no use for it if you could, but it explains why Google shipped two models today instead of one.

Google says it is already running Flash Cyber on its own code. It reports that the Chrome Security team got "2.6 times more correct patches" to Chrome vulnerabilities than from much larger commercial models, and that its Cloud Vulnerability Research team used the model to find a critical vulnerability "in less than 2 hours" for work that "usually takes months". On CWE-Bench, an external patching benchmark, Google reports a pass@1 of 47.2% against a leading frontier model at 47.8%, at what it says is a significantly lower cost.

Access is by application to the Fairwind Program, which Google announced the same day for "trusted government authorities, as well as critical infrastructure operators and software maintainers". Google also says Flash Cyber "ships with a more permissive set of mitigations for cybersecurity", which is the reason it is gated: a model trained to find exploitable flaws is useful to both sides.

There is one detail in the safety section that does touch ordinary users. Google says the 3.8 models "made a significant leap in prompt injection robustness as measured by Gray Swan". Prompt injection is what happens when a web page or email contains hidden instructions aimed at your assistant rather than at you, and it is the main thing that can go wrong when an AI assistant browses on your behalf. Google publishes no number for the improvement, so take it as a direction, not a measurement.

Should You Care?

Not much, unless you already pay Google. Here is the short version by plan:

• Free plan: nothing changed on September 2, but something did on October 9, 2026. Google's plan page moved the $0 tier from 3.6 Flash to 3.5 Flash-Lite and put 3.6 Flash behind $4.99 AI Plus. See Gemini 3.6 Flash, explained.
• Google AI Plus ($4.99): Google did not name you. Assume nothing changed until the plan page says otherwise.
• Google AI Pro or Ultra: your Gemini app, AI Mode and Google Sheets got a better model at no extra cost and with nothing to switch on. Point it at the long, multi-step jobs that failed before, which is where Google's own numbers improve.
• Anyone weighing an upgrade: a model swap you cannot see is not a reason to start paying $19.99 a month. Our guide to which AI chatbot is worth paying for works through the cases where the paid tier earns it, and the best free AI chatbot in 2026 covers what $0 still buys.

The bottom line: Gemini 3.8 Flash is a genuine upgrade aimed at coding and agent work, and for the first time since July a new Flash model reaches the Gemini app that ordinary people use. But it reaches it behind a subscription. Google names AI Pro and Ultra and nobody else, publishes no release note, and leaves its own plan page saying 3.6 Flash. If you are not paying, the honest answer is that today's model launch is not for you either.

Corrections and Updates

October 19, 2026: Google deleted the model version numbers from its plan page, so the wording this page quoted no longer appears there. The $0 card read "Access to 3.5 Flash-Lite" in the Internet Archive capture of the subscriptions page taken October 13, 2026 at 15:18 GMT, and reads "Access to our Flash-Lite model" live today; the $4.99 card went from "Access to 3.6 Flash" to "Access to our Flash model", and "3.5 Flash-Lite" and "3.6 Flash" now return zero occurrences on the whole page. Every tier lost its version number in the same rewrite and the models help page names them without versions too, so no Google consumer surface now states which Flash version a given plan gets. That is a change of label, not of plan: nothing read today says the free plan's model moved again, and this page does not claim it did. Quotes here are now dated to the capture that contains them.

October 19, 2026: two places here said Google's AI Mode help article "still describes only Gemini 3 Pro" as the selectable model. It does not, and has not since mid-October 2026. The FAQ answer and the availability table's AI Mode row both carried that in the present tense. Read live on October 19, that article documents Gemini 4 Argon in AI Mode and refers to the alternative only as the "Auto" or "Fast" model; "Gemini 3" returns zero occurrences on the page. Neither claim was load-bearing for what 3.8 Flash is or who gets it -- Google's one sentence about 3.8 Flash in AI Mode is unchanged -- but both were false as written and both are surfaces an answer engine quotes.

October 13, 2026: this page said the newest Gemini release note was "still 2026.08.19" and cited help article 16275805 for the app's model list. Both are wrong. Three release notes have landed since -- 2026.09.10, 2026.09.30 and 2026.10.13, the last of them Gemini 4 Argon -- and Google moved the model list off 16275805 into a new article, About Gemini models, which lists Flash-Lite, Flash, Pro and Argon. The Key Takeaway, the switching FAQ and the surface table now say so. This page's verdict is unchanged: "3.8" still appears on none of those surfaces, so there is still nothing for you to pick.

October 11, 2026: this page said the free plan was "still on 3.6 Flash". Google moved it to 3.5 Flash-Lite on October 9, 2026 and announced nothing. Its subscriptions page read "Access to 3.6 Flash" on the $0 tier in the Wayback capture of October 9 at 07:14 GMT and "Access to 3.5 Flash-Lite" in the capture of the same day at 21:37 GMT, with 3.6 Flash moved to the $4.99 Google AI Plus tier; the live page read today agrees. The Gemini release notes have no entry for it, newest 2026.09.30. Every free-plan row on this page now says 3.5 Flash-Lite.