What is Gemini 3.6 Flash?
Gemini 3.6 Flash is a Google text model with multimodal input, released on 21 July 2026 as the third Flash release of the Gemini 3 generation. Its API model code is gemini-3.6-flash, Google lists it as Stable rather than preview, and the documented limits are a 1,048,576-token input and a 65,536-token output. The documented inputs are text, images, video, audio and PDF; the documented output is text. Google states plainly that the model is built on Gemini 3.5 Flash and points readers to that model card for its architecture, training data, hardware and software, so this is a refinement of the previous Flash rather than a new line. Two things about it are easy to get wrong, and both change how you would use it. The first is money. Google publishes four rate tables for this model and every one of them quotes two figures with two dates: on the Standard tier an input costs $0.75 per million tokens through December 31, 2026 and $1.50 starting January 1, 2027, while an output costs $3.75 and then $7.50. That dated pair is specific to this generation. The same pricing page gives Gemini 3.5 Flash a single undated figure of $1.50 in and $9.00 out, with no end date anywhere in its four tables, so two models in the same family are not priced the same way even though one is built from the other. The second mistake runs the other direction. Google’s pricing page describes this model, in the sentence that names it, as “our previous generation Flash model”; the API model page lists it as Stable, and a banner at the top of that page points at Gemini 3.8 Flash. Those are different claims from the same provider, and this page reports each by source rather than choosing between them. What is not in doubt is the withdrawal status: the deprecations page carries a row for gemini-3.6-flash with a release date of 21 July 2026 and an empty shutdown-date column, and it states no shutdown date. Two later Flash models have shipped since, and that same table points the older gemini-3-flash-preview at this model as its recommended replacement.
What Gemini 3.6 Flash is worth using for
Read the dated price window before you budget anything
This is the part of the model that most pages about it get wrong, and the error is mechanical rather than subtle. Google’s pricing page for Gemini 3.6 Flash prints a price and a date on the same line, four times over. Standard input is $0.75 per million tokens through December 31, 2026 and $1.50 starting January 1, 2027; Standard output is $3.75 and then $7.50. Batch and Flex sit at $0.375 and $1.875 in the same two-step pattern, and Priority, the dearest tier, runs $1.35 and $6.75. Every tier carries both dates. A large share of the pages that quote a price for this model quote only $1.50 and $7.50, which is the second of those two figures, and present it as though it were the current rate. Google’s own model card and launch announcement do the same thing, stating $1.50 and $7.50 without dates. So there are two official readings in circulation and they are not interchangeable: the pricing page is the one with dates attached, and it is the one to budget from, because it is the page that will still be right after the window closes. In practice that means a cost estimate built this year should use the lower pair and should carry the end date with it, because the higher pair is already published as the next state rather than as a surprise. It also means a comparison against Gemini 3.5 Flash has to be read carefully, since that model has no dated clause at all: its $1.50 input is not a promotional number about to rise.
I want to estimate what a long-running job on this model will cost next year. Ask me how many input and output tokens I expect per month. Then show me the two published price states side by side with their dates, tell me which one applies to which period, and flag every place where an article or a comparison page would give me the wrong number.
Separate “previous generation” from “withdrawn”
Two different Google pages answer two different questions about this model, and reading one as an answer to the other is the most common mistake in circulation. The pricing page answers where the model sits in the line, and its answer is blunt: introducing the section, it calls Gemini 3.6 Flash “our previous generation Flash model, balancing speed and multimodal capabilities across general agentic and everyday tasks”. Since that sentence was written, Google has released Gemini 3.7 Flash on 13 August 2026 and Gemini 3.8 Flash on 2 September 2026, and the API model page now carries a banner advertising 3.8 Flash. So on the Flash line this model is two releases behind, and the provider says so itself. The deprecations page answers a different question, whether the model is going away, and its answer is equally clear: the row for gemini-3.6-flash has a release date and an empty shutdown-date column, and the page states that no shutdown date is announced. Two rows away, Gemini 3.1 Flash-Lite carries a real shutdown date of 7 May 2027 and a named replacement, so the table genuinely distinguishes the two states rather than leaving the column blank for everything. Meanwhile the recommendation runs the other way as well: the table points the older preview model gemini-3-flash-preview at gemini-3.6-flash as its replacement, which makes this model the destination for at least one migration rather than only a waypoint on the way to 3.8. The sentence that survives all of this has two halves, and this page writes both wherever the question comes up: Gemini 3.6 Flash is a stable, still-served model that two newer releases have overtaken.
I read that Gemini 3.6 Flash is an older model and that a newer one replaces it. Walk me through how to check whether it is merely superseded or actually being retired: which official pages to open, what to look for on each, and what result on each would change the conclusion. Then tell me exactly what would have to be true for a retirement claim to be correct.
Read the model-card table with its caveats attached
Google’s model card for this release publishes one evaluation table, and the column describing this model is the first one. Ten rows of it are reproduced further down this page, and every one moves in the same direction against the model this release was built on, Gemini 3.5 Flash: SWE-Bench Pro reads 58.7% against 55.1%, DeepSWE v1.1 reads 49% against 37%, MLE-Bench 63.9% against 49.7%, Terminal-bench 2.1 78.0% against 76.2%, and GDM-MRCR v2 at 128k reads 91.8% against 77.3%. That much is unambiguous. What the table does not support is a claim that this model leads everything on the page, and the same card makes that clear without any outside source. On SWE-Bench Pro it also prints GPT-5.6 Luna at 62.7%, Grok 4.5 at 64.7% and Claude Sonnet 5 at 63.2%, all above this model; on GDPVal-AA v2, where this model reads 1421 on the Elo scale, Claude Sonnet 5 reads 1607, GPT-5.6 Luna 1584 and Grok 4.5 1535. Two of the ten rows need one more caution, because neither is a single measurement: CharXiv Reasoning appears twice, once with tools and once without, and GDM-MRCR v2 appears twice, as a 128k average and as a 1M pointwise reading. Each pair is two scoring scales of one benchmark rather than two benchmarks, and the row labels carry the scale so the two cannot be averaged together. The card also links its methodology out to a separate page rather than stating it inline, and the card prints two price rows in the same table, which this page deliberately keeps out of the comparison and reports in the pricing section instead.
Here is an official benchmark table with several model columns. Compare the first column with the column I name, give me the size of every gap, and then point out the rows where the first column is not the highest. For each of those rows, tell me which model leads and by how much.
Keep the official capabilities apart from this site’s switches
Two sets of statements describe this model and they answer different questions. On the official side there is a 1,048,576-token input limit, a 65,536-token output limit, and a capability list that includes caching, code execution, file search, function calling, structured outputs, URL context, search grounding, Google Maps grounding, thinking, and computer use marked as a preview. On this site’s side the catalog entry behind the page declares vision and web search as supported, so images, PDFs and the search switch are available, and it declares reasoning as unsupported, so the composer offers no way to choose a thinking level even though Google lists thinking among the model’s capabilities. That asymmetry is worth stating plainly, because here it is the reasoning control that is missing and the search switch that is present. Attachments accept JPEG, PNG, WebP and PDF files up to 24 MB, and a PDF is converted to text before the model sees it, so a scan the extractor cannot read comes back as an error instead of an empty answer; audio and video are documented inputs of the model but cannot be uploaded here. Each reply is requested with at most 4,096 output tokens, which is a setting of this site and not a property of the model, and because the catalog entry publishes no context length the interface displays no context figure at all. Read the specification cards as Google describing Gemini 3.6 Flash, and the controls in the chat box as what this site actually switches on. One more detail belongs on the official side of that line: the catalog entry here publishes no price for web search even though the switch is available, so nothing on this page should be read as the cost of one.
I want to use this chat box for a task that needs a lot of source material. Ask me what the task is before answering. Then tell me which parts of it are limited by this interface rather than by the model, which control I do have here that I might not expect, and how to split the work around the per-reply ceiling.
Official Gemini 3.6 Flash specifications
- Developer
- Model ID
- gemini-3.6-flash
- Official input limit
- 1,048,576 tokens
- Official output limit
- 65,536 tokens
- Official inputs
- Text, images, video, audio and PDF
- Output
- Text
Gemini 3.6 Flash and Gemini 3.5 Flash side by side
Every value below comes from one table, the evaluation table inside Google’s Gemini 3.6 Flash model card, checked on 24 September 2026, so a single official page carries both columns. The left column is this page’s model and the right column is Gemini 3.5 Flash, which the card itself names as the model this release is built on; it is also the comparison most often searched for, and the only one where both columns sit in the same official table. Rows are matched by benchmark name, never by position. The card’s table also prints two price rows, and they are deliberately left out of this comparison: the pricing page publishes dated figures for this model while the card states $1.50 and $7.50 without dates, so prices are reported in the pricing section by source and are not mixed into a benchmark table. Three things have to travel with the numbers. All ten rows move in this model’s favour, which makes a clean sweep easy to imply and wrong to claim, because the same card prints other models above this one on SWE-Bench Pro, GDPVal-AA v2 and Terminal-bench 2.1. CharXiv Reasoning and GDM-MRCR v2 each appear twice, as two scoring scales of one measurement, and their row labels carry the scale so the two cannot be read together. And the card links its evaluation methodology out to a separate page rather than stating it inline. Read this as a provider comparing two of its own releases, not as a promise about the quality or speed of a conversation on this site.
| Provider benchmark | Gemini 3.6 Flash | Gemini 3.5 Flash |
|---|---|---|
| SWE-Bench Pro (Public) | 58.7% | 55.1% |
| DeepSWE v1.1 | 49% | 37% |
| Terminal-bench 2.1 | 78.0% | 76.2% |
| MLE-Bench | 63.9% | 49.7% |
| GDPVal-AA v2 | 1421 | 1349 |
| OSWorld-Verified | 83.0% | 78.4% |
| CharXiv Reasoning (no tools) | 85.2% | 84.2% |
| CharXiv Reasoning (with tools) | 89.4% | 84.9% |
| GDM-MRCR v2 (8-needle) · 128k avg | 91.8% | 77.3% |
| GDM-MRCR v2 (8-needle) · 1M pointwise | 54.0% | 26.6% |
How to use Gemini 3.6 Flash online
Choose the model
This page opens with Gemini 3.6 Flash selected. The model navigation on the left and the selector in the chat box stay in sync, and switching models keeps the current conversation. Start a new chat when you want a clean context.
Write the task and the output format
State the goal, the constraints and the shape of the answer you want. This box exposes no reasoning-level control for this model, so a tightly scoped instruction is the main lever you have; the search switch, by contrast, is available here. Any prompt in this guide can be sent as it stands or edited first.
Sign in, send and check the cost
Sending a message needs an account with available points. The composer reads a reference price from the live model catalog, and a finished request reports what it actually cost. Follow up with corrections rather than starting over.
What Gemini 3.6 Flash costs
Google publishes four rate tables for this model, all quoted per million tokens in US dollars, and all four attach two dates to the same two figures. On the Standard tier an input costs $0.75 through December 31, 2026 and $1.50 starting January 1, 2027, and an output costs $3.75 and then $7.50, with the output figure stated to include thinking tokens; a cache read costs $0.075 and then $0.15, and cached storage is billed separately at $0.50 per million tokens per hour through the same date, rising to $1.00. Batch and Flex both sit at $0.375 for input and $1.875 for output, with cache reads of $0.0375 rising to $0.075, and Priority is the dearest tier at $1.35 for input and $6.75 for output with a cache read of $0.135, each rising to $2.70, $13.50 and $0.27 on the same date. Two rows are shared across the tiers: Grounding with Google Search is unavailable on the free tier and on paid tiers includes 5,000 free search requests per month shared across the Gemini 3.x family before being billed at $14 per 1,000 requests, and Grounding with Google Maps includes 5,000 prompts per month at the same rate thereafter. The free tier is worth reading per tier rather than as a whole, because it reads “free of charge” on Standard and Priority and “not available” on Batch and Flex. This dated pair is not the only official statement of price. Google’s model card and its launch announcement both give $1.50 and $7.50 without dates, so the two readings are reported here by source: the pricing page carries the dates, the other pages do not. None of that is this site’s checkout price. Here you pay in points, and as of the information-check date the catalog entry lists 91 points per million input tokens, 455 points per million output tokens and 9 points per million cached tokens; it publishes no separate price for a web search even though search is available for this model here.
View points and plans ↗Gemini 3.6 Flash FAQ
Is Gemini 3.6 Flash deprecated or scheduled to shut down?
Google has published no shutdown date for it. The official deprecations page carries a row for gemini-3.6-flash with a release date of 21 July 2026 and an empty shutdown-date column, and its recommended-replacement column is empty as well. That is a different arrangement from Gemini 3.1 Flash-Lite, further down the same table, which carries a shutdown date of 7 May 2027 and names a replacement. Being superseded and being withdrawn are separate things, and for this model only the first is true: Google has since released Gemini 3.7 Flash and Gemini 3.8 Flash, and the model is still served.
What does Gemini 3.6 Flash cost, and does the price have an end date?
Both questions are answered on one page, and the answer to the second one is yes. Google’s pricing page quotes every tier as a pair of figures with a pair of dates. Standard input is $0.75 per million tokens through December 31, 2026 and $1.50 starting January 1, 2027; Standard output is $3.75 and then $7.50. Batch and Flex sit at $0.375 and $1.875, and Priority at $1.35 and $6.75, each with the same two-step pattern. That dated arrangement is specific to this generation: the same page gives Gemini 3.5 Flash a single undated figure, so its $1.50 input is not a promotional number about to rise. Google’s model card and launch announcement also state $1.50 and $7.50, but without dates, which is where the common error of quoting the later pair as the current rate comes from.
How does Gemini 3.6 Flash compare with Gemini 3.5 Flash?
The model card for this release publishes one evaluation table, Gemini 3.5 Flash is one of its columns, and that is why the table above uses it: it is the only comparison where a single official page proves both columns. Across the ten rows shown, the newer model is ahead on every one, with the largest gaps in long-horizon coding and machine-learning engineering work — DeepSWE v1.1 at 49% against 37%, MLE-Bench at 63.9% against 49.7%, and GDM-MRCR v2 at 128k at 91.8% against 77.3%. The provider’s own launch post adds a token-efficiency figure for the same pair: on the Artificial Analysis Index, this model consumes 17% fewer output tokens than Gemini 3.5 Flash, rising to as much as 65% on DeepSWE. Note what that 17% measures — it is output token usage, not a speed or a price difference.
Which models replaced Gemini 3.6 Flash, and when were they released?
Google’s deprecations table gives the dates directly. Gemini 3.6 Flash was released on 21 July 2026, and Gemini 3.5 Flash-Lite shares that date. Gemini 3.7 Flash followed on 13 August 2026 and Gemini 3.8 Flash on 2 September 2026. Neither later generation carries a shutdown date either. The pricing page states the position this model holds now, describing it in the sentence that names it as “our previous generation Flash model”, while the API model page still lists it as Stable and a banner on that page points at 3.8 Flash. The recommendation column of the same deprecations table runs the other way as well, pointing the older gemini-3-flash-preview at this model as its replacement.
Is Gemini 3.6 Flash a new architecture?
No, and Google says so directly. The model card’s description of dependencies reads that Gemini 3.6 Flash is built on Gemini 3.5 Flash, and the sections covering architecture, training dataset, training-data processing, hardware and software all refer readers to the Gemini 3.5 Flash model card rather than restating anything. What the card claims for this release is better coding, knowledge-work and multimodal performance with better token efficiency than its predecessor, which is a refinement claim rather than an architecture claim. The documented limits changed in no way that distinguishes the two: both are published with a 1,048,576-token input and a 65,536-token output.
Can Gemini 3.6 Flash read images and PDFs here?
Yes to both, and this is a point where the official specification and the site controls line up instead of differing. Google documents text, image, video, audio and PDF as inputs and text as the output, and the catalog entry behind this page enables vision. The composer therefore accepts JPEG, PNG, WebP and PDF files up to 24 MB. A PDF is not handed to the model as a document: this site extracts its text first, so a scanned page the extractor cannot read comes back as an error rather than an empty answer, and complex layouts may need a clearer image or manually pasted text. Audio and video uploads are not available here even though the model documents them as inputs.
Why is there no thinking-level selector in this chat box?
Because the catalog entry behind this page declares reasoning as unsupported, so the composer has no control to show. The official model differs here: Google lists thinking among the capabilities of Gemini 3.6 Flash. That is a gap between the official model and this interface rather than a property of the model, and it is stated here instead of left to be discovered. It is also worth noting that this page is the mirror image of the usual pattern in one respect: web search is switched on here, so the reasoning control is the one that is missing. And because this model is based on Gemini 3.5 Flash, anything you read about the previous model’s thinking levels describes the same underlying line.
How long can an input or an answer be?
This site requests at most 4,096 output tokens per reply, which is a workspace setting and not a limit of the model. Google documents a 1,048,576-token input limit and a 65,536-token maximum output for the API. Because the catalog entry behind this page publishes no context length, the interface shows no context figure, so the official number is an API specification rather than a description of this page. Long deliverables are best asked for in parts, with the shared context carried forward between them. One caveat is worth carrying: the model card scores the same long-context benchmark twice, at 91.8% for a 128k average and 54.0% at the full 1M pointwise reading, so a larger context is not automatically a better one.
Sources and model information
Information checked
Start with the task you run most often
Give it the goal, the constraints and the format you need, and keep the inputs within what this chat box can actually take.
Chat with Gemini 3.6 Flash ↑