Gemini

Google's model family, available through the Gemini app and the Gemini API. Gemini 4 Carbon, launched on 10 November 2026, is Google's strongest model; Gemini 4 Argon is the faster, lower-cost Gemini 4 option. Several Flash and Flash-Lite models cover high-volume work, some of them new or in preview.

Summary is our draft

Sources checked: 11 Nov 2026
Our testing: Not started

Current lineup

Gemini 4 Carbon
Our strongest reasoning model
gemini-4-carbon
Gemini 4 Argon
Our efficient Gemini 4 frontier reasoning model, delivering faster responses and lower token spend per turn
gemini-4-argon
Gemini 3.8 FlashNew
Our most intelligent Flash model, engineered for long-horizon software engineering, autonomous agents, and complex enterprise workflows.
gemini-3.8-flash
Gemini 3.6 FlashStable
Our previous-generation Flash model, balancing speed and multimodal capabilities across general agentic and everyday tasks.
gemini-3.6-flash
Gemini 3.5 Flash-LiteStable
No provider description captured
gemini-3.5-flash-lite
Gemini 3.1 Pro
No provider description captured
gemini-3.1-pro-preview

Verified 12 Nov 2026Google AI Studio: Gemini models (opens in a new tab)

Which one for you

Each line says whether it comes from the provider or is our own judgment from the checked facts. None of it is from our own testing yet.

Work involving audio, video or PDFs
Gemini 4 Argon accepts text, images, audio, video and PDF, and Carbon accepts text, images, audio and video: the widest input lists of the flagships we profile.
Our judgment
Choosing between the two Gemini 4 models
Google calls Carbon its strongest model and Argon the faster, lower-cost one per turn. Start with Argon and move to Carbon if results fall short. Per-token API prices are the same for both, during the introductory period and from 1 Dec 2026. Google's Carbon model card says Carbon uses more thinking tokens than Argon, so in our judgment Argon will usually cost less per completed task.
Our judgment
Anything you'll rely on in production
Prefer models Google marks Stable. Preview and New models can change or be withdrawn.
Our judgment
The cheapest paid app plan
Google AI Plus at $4.99 a month is the lowest-priced paid plan we could verify across the providers profiled here (US prices).
Our judgment

Key facts

Context window: 1M tokens (Gemini 4 Argon); Source, 12 Nov 2026 (opens in a new tab)
Context window (Carbon): Up to 1M tokens of input and output combined; Source, 12 Nov 2026 (opens in a new tab)
Max output: Gemini 4 Argon: 64K tokens (65,536) by default; up to 1M with long decode, which needs background execution and shares the 1M window with your input. Gemini 4 Carbon: Google's model card gives up to 1M tokens of input and output combined, with no separate output limit; Source, 12 Nov 2026 (opens in a new tab)
Inputs (Argon): Text, images, audio, video and PDF; Source, 12 Nov 2026 (opens in a new tab)
Inputs (Carbon): Text, images, audio and video. Google's model card doesn't list PDF; Source, 12 Nov 2026 (opens in a new tab)
Knowledge cutoff: March 2026 (Gemini 4 Carbon); Source, 12 Nov 2026 (opens in a new tab)
Other models: Listed on each model's own page: not yet checked
Gemini 2.5 models: Google says they are not deprecated and will continue to be served until further notice. For new projects it points to 3.5 Flash-Lite or 3.8 Flash; Source, 12 Nov 2026 (opens in a new tab)

By task

We haven't run our own tests yet, so this shows only what the provider claims and the test every family will get. Results will appear here, labelled as ours.

Coding
The provider saysNo specific claim captured
The test we’ll runRefactor a 400-line TypeScript module with an existing test suite, then fix one failing test.
Writing
The provider saysNo specific claim captured
The test we’ll runTurn a rough 600-word draft into a 300-word brief without losing any of its facts.
Research
The provider saysNo specific claim captured
The test we’ll runSummarise three supplied PDFs and cite the page for every claim.
Reasoning
The provider saysNo specific claim captured
The test we’ll runPlan a five-step project with dependencies and explain which step blocks which.

Pricing and access

Pricing

Gemini 4 Carbon: $2 in / $10 out per million tokens (introductory); $4 in / $20 out after 30 Nov 2026Source, 12 Nov 2026 (opens in a new tab)
Gemini 4 Argon: $2 in / $10 out per million tokens (introductory, through 30 Nov 2026); $4 in / $20 out from 1 Dec 2026. Not on the free tierSource, 12 Nov 2026 (opens in a new tab)
Flash and Flash-Lite: Per-model API pricing (Google's pricing page doesn't yet list these models): not yet checked
App plans (US): Free; Google AI Plus $4.99 a month; AI Pro $19.99 a month; AI Ultra from $99.99 a monthSource, 12 Nov 2026 (opens in a new tab)

Where you can use it

Developer platforms: Gemini API (Google AI Studio)Source, 12 Nov 2026 (opens in a new tab)
Consumer apps: Gemini app; Google says Gemini 4 Argon is included in AI Plus, Pro and UltraSource, 12 Nov 2026 (opens in a new tab)
Gemini 4 Carbon in the app: Rolling out to Google AI Ultra subscribersSource, 12 Nov 2026 (opens in a new tab)
Gemini 4 Carbon in the API: Google AI Studio and Gemini for business, for paying developers

Before you choose

Preview and New models can change or be withdrawn. Prefer models marked Stable for anything you rely on.
Gemini 4 prices are introductory until 30 Nov 2026 and double from 1 Dec 2026, according to Google's Carbon launch post and Argon documentation. We'll re-check both after 1 Dec.
Google's API pricing page doesn't yet list Gemini 4 or the newer Flash models. Gemini 4 prices come from the launch post and Google's Gemini 4 Argon documentation; Flash pricing is shown as not yet checked.

Our draft notes

Put it to work

What changed

12 Nov 2026Pre-launch re-check. Google's AI Studio models index no longer lists Gemini 4 Carbon, so Carbon's context window and inputs are now cited to its model card and shown separately from Argon's. The card lists text, images, audio and video, not PDF, so we no longer say Carbon accepts PDF. Gemini 2.5 wording updated to Google's current text.
11 Nov 2026Correction: we had listed max output as "65K / 1M tokens" for both Gemini 4 models, which ran the default limit and long decode together. Google's Argon documentation gives 64K by default and up to 1M only with long decode; Carbon's model card gives a combined 1M input and output. We also removed a line saying Carbon would cost more after 30 Nov: per-token prices are the same for both.
11 Nov 2026Correction: we had said Gemini 4 Argon's price after its introductory period wasn't stated. Google's Argon documentation says it rises to $4 in / $20 out per million tokens from 1 Dec 2026, the same as Carbon. Pricing line and sources updated.
11 Nov 2026Added Gemini 4 Carbon (launched 10 Nov 2026) with its introductory API price, app availability and knowledge cutoff. Google now describes Argon as its efficient Gemini 4 model rather than its flagship.
11 Nov 2026Removed the note that Gemini 2.5 Flash would be rerouted from 1 Dec 2026. Google's page now says Gemini 2.5 models are not deprecated.
11 Nov 2026Added provider descriptions, Gemini 4 Argon's input types and US app plan prices.

Sources

Google AI Studio: Gemini models (opens in a new tab)Checked 12 Nov 2026
Google: Gemini subscriptions (opens in a new tab)Checked 12 Nov 2026
Google: Gemini 4 Carbon now available (10 Nov 2026) (opens in a new tab)Checked 12 Nov 2026
Google AI for Developers: Gemini 4 Argon (latest model) pricing (opens in a new tab)Checked 12 Nov 2026
Google DeepMind: Gemini 4 Carbon model card (opens in a new tab)Checked 12 Nov 2026

Current lineup

Which one for you

Key facts

By task

Coding

Writing

Research

Reasoning