Gemini models in 2026: 3.8 Flash, 3.1 Pro, Live, Flash-Lite, Nano Banana, and app model names

Gemini now has two naming layers: simple app labels and explicit developer model IDs. In the Gemini app, users see Flash-Lite, Flash, and Pro; in the API, developers choose specific models such as Gemini 3.8 Flash and Gemini 3.1 Pro Preview.

The September 2026 lineup also includes Gemini 3.8 Live and Live Extended Thinking for realtime voice, Flash-Lite variants for low-cost workloads, and specialized image, video, transcription, and media-generation models.

··········

THE CURRENT GEMINI MODEL MAP

Gemini's consumer app no longer exposes every developer model ID. App labels are intentionally simpler, while the API keeps explicit versioned endpoints for reproducibility, pricing and migration.

........

Model / label	Where it appears	Current role	Context / output
Gemini Pro / Gemini 3.1 Pro Preview	Gemini app / API	Highest general-purpose reasoning tier in the current Gemini 3 app/API stack	1M / 64K in API
Gemini 3.8 Flash	Gemini API and AI Studio	GA workhorse for coding, agents and complex enterprise workflows	1M / 64K
Gemini Flash	Gemini app	Balanced speed and reasoning; underlying version can change	Plan-dependent app context
Gemini Flash-Lite	Gemini app	Fast, efficient everyday tasks	Plan-dependent app context
Gemini 3.8 Live / Extended Thinking	Gemini API and AI Studio	Realtime audio-to-audio agents and live dialogue	Realtime-specific
Nano Banana 2 / Pro	Gemini API and Google products	Image generation and editing	Specialized image models

........

··········

GEMINI 3.8 FLASH IS THE CURRENT GA FLASH REFERENCE

Gemini 3.8 Flash became generally available in September 2026. Google positions it for long-horizon software engineering, autonomous agents and complex enterprise workflows, with 1,048,576 input tokens and 65,536 maximum output tokens.

··········

GEMINI 3.1 PRO REMAINS THE CURRENT PRO API REFERENCE

The API currently exposes gemini-3.1-pro-preview for complex multimodal reasoning, software engineering and agentic workflows. It accepts text, images, video, audio and PDFs and supports a 1M-token input window with 64K output.

··········

FLASH-LITE, FLASH, AND PRO IN THE APP ARE PRODUCT LABELS

The Gemini app presents three user-facing model categories: Flash-Lite, Flash and Pro. Those names should not be treated as permanent API IDs because Google can update the underlying implementation while preserving the app-level product label.

··········

LIVE MODELS FORM A SEPARATE REALTIME BRANCH

Gemini 3.8 Live and 3.8 Live Extended Thinking target low-latency voice agents. They are priced and configured differently from text-centric Flash and Pro models and belong in a realtime architecture rather than a standard text-model comparison.

··········

IMAGE AND MEDIA MODELS ARE PART OF THE GEMINI ECOSYSTEM, NOT EXTRA CHAT MODELS

Nano Banana 2, Nano Banana Pro, Gemini Omni Flash, Veo and related media systems use separate endpoints and pricing. They should be listed as specialized Gemini-family models rather than ordinary chat choices.

··········

LEGACY DOES NOT ALWAYS MEAN SHUT DOWN

Google's deprecation page distinguishes active, deprecated and shutdown endpoints. Gemini 3 Flash Preview is now described as a legacy Flash model and Google recommends newer Flash generations, but a legacy label is not the same as an already-disabled endpoint.

··········

HOW TO CHOOSE THE RIGHT GEMINI MODEL

For the Gemini app, choose between Flash-Lite, Flash and Pro based on speed and reasoning needs. For production API work, choose an explicit model ID, verify its deprecation status, calculate token economics and keep migration logic separate from consumer-app labels.

··········

FOLLOW US FOR MORE

··········

DATA STUDIOS

··········

datastudios.org