Learn about supported models

For mobile and web apps, the Firebase AI Logic SDKs let you interact with the supported Gemini models directly from your app.

Gemini models are considered multimodal because they're capable of processing and even generating multiple modalities, including text, code, PDFs, images, video, and audio.

Also, review our FAQ about all the models that Firebase AI Logic supports and does not support.

FLAGSHIP FRONTIER

gemini-4-argon

Advanced reasoning and rich multimodal understanding to handle complex use cases and intelligent in-app workflows. (billing required)

FAST AND INTELLIGENT

gemini-3.8-flash

Frontier-class performance model that rivals larger models at a fraction of the cost. (billing not required)

ULTRA FAST

gemini-3.5-flash-lite

High-volume, cost-sensitive workhorse model with the performance and quality of the Gemini 3 series. (billing not required)

IMAGE GENERATION

gemini-nano-banana-2.1

Fast, cost-efficient workhorse model for high-quality image generation and conversational editing. (billing required)

General-use models

Go to tables with model details

gemini-4-argon

Advanced reasoning and rich multimodal understanding to handle complex use cases and intelligent in-app workflows. (billing required)

gemini-3.8-flash

Frontier-class performance model that rivals larger models at a fraction of the cost. (billing not required)

gemini-3.5-flash-lite

High-volume, cost-sensitive workhorse model with the performance and quality of the Gemini 3 series. (billing not required)

Image-generating models (Nano Banana)

Go to tables with model details

gemini-nano-banana-2.1

Primary workhorse model for image generation and conversational editing. Fast and cost-efficient, with improved visual quality, text rendering, and multi-turn consistency. (billing required)

gemini-3.1-flash-image

High-efficiency model optimized for fast, reliable image generation across production workloads. (billing required)

gemini-3.1-flash-lite-image

Ultra-low latency and cost-effective model, designed for high-volume, real-time interactive use cases. (billing required)

gemini-3-pro-image

State-of-the-art model for complex image prompts, structured output, and maximum visual fidelity. (billing required)

Audio-generating models

Text-to-speech (TTS) models

You can generate speech from text input with Gemini TTS models.

Go to tables with model details

gemini-3.1-flash-tts-preview

Powerful, low-latency speech generation from text input. (billing not required)

Live API models

You can generate bidirectional streamed audio with models that support the Gemini Live API.

Go to tables with model details

Gemini Developer API:
gemini-3.1-flash-live-preview

Agent Platform Gemini API:
not supported

Enables low-latency, real-time voice and video interactions with a Gemini model that is bidirectional. (billing not required)

Gemini Developer API:
gemini-2.5-flash-native-audio-preview-12-2025

Agent Platform Gemini API:
gemini-live-2.5-flash-native-audio

Enables low-latency, real-time voice and video interactions with a Gemini model that is bidirectional. (billing not required)


The remainder of this page provides detailed information about the models supported by Firebase AI Logic.

  • Compare models:

    • Supported input and output
    • High-level comparison of the supported capabilities
    • Specifications and limitations, for example max input tokens or max length of input video

  • Description of how models are versioned, specifically their stable, preview, and experimental versions

  • Lists of available model names to include in your code during initialization

  • Lists of supported languages for the models

At the bottom of this page, you can view detailed information about previous generation models.


Compare models

Each model has different capabilities to support various use cases. Note that each of tables in this section describe each model when used with Firebase AI Logic. Each model might have additional capabilities that aren't available when using our SDKs.

If you can't find the information you're looking for in the following sub-sections, you can find even more information in your chosen API provider documentation: Gemini Developer API or Agent Platform Gemini API (formerly Vertex AI).

Supported input and output

The following table lists the supported input and output types when using each model with Firebase AI Logic.

To learn about supported file types, see Supported input files and requirements.

Gemini 4
Argon 		Gemini 3.x
Flash, Flash‑Lite 		Gemini
Nano Banana
2, 2.1 		Gemini
Nano Banana
2 Lite 		Gemini
Nano Banana
Pro 		Gemini 3.x
Flash
TTS 		Gemini
Live
Input types
Text
Code
Documents
(PDFs or plain-text)
Images
Video
Audio
Output types
Text
Text (streaming)
Code
Structured output
(like JSON)
Images
Video
Audio

Supported capabilities and features

The following table lists the supported capabilities and features when using each model with Firebase AI Logic.

Gemini 4
Argon 		Gemini 3.x
Flash, Flash‑Lite 		Gemini
Nano Banana
2, 2.1 		Gemini
Nano Banana
2 Lite 		Gemini
Nano Banana
Pro 		Gemini 3.x
Flash
TTS 		Gemini
Live
Thinking
Generate text from text-only or multimodal inputs interleaved or as part of image interleaved or as part of image interleaved or as part of image transcription of audio
Generate images
Edit images
Generate audio speech streamed audio
Generate structured output
(like JSON)
Analyze documents
(PDFs or plain-text)
(text-output | image-output)
Analyze images
(text-output | image-output)
Analyze video
(text-output | image-output) 		streamed video
Analyze audio streamed audio
Multi-turn chat
Bidirectional multimodal streaming
Supported tools
Function calling
Code execution
URL context
Grounding with Google Search
Grounding with Google Maps

Specifications and limitations

The following table lists the specifications and limitations when using each model with Firebase AI Logic.

Property Gemini 4
Argon 		Gemini 3.x
Flash, Flash‑Lite 		Gemini
Nano Banana
2, 2.1 		Gemini
Nano Banana
2 Lite 		Gemini
Nano Banana
Pro 		Gemini 3.x
Flash
TTS 		Gemini
Live
Input token limit * 1,048,576
tokens		 1,048,576
tokens		 131,072
tokens		 65,536
tokens		 65,536
tokens		 8,192
tokens		 see docs
Output token limit * 32,768 tokens (non‑streaming)
65,536 tokens (streaming) **		 65,536
tokens		 32,768
tokens		 4,096
tokens		 32,768
tokens		 16,384
tokens
PDFs (per request)
Max number
of input PDF files ***		 900 files 900 files 14 files 14 files 14 files --- see docs
Max number
of pages
per input PDF file ***		 900 pages 900 pages 14 pages 14 pages 14 pages ---
Max size
per input PDF file		 50 MB 50 MB 50 MB 50 MB 50 MB ---
Images (per request)
Max number
of input images		 1,000 images 1,000 images 14 images 14 images 14 images --- see docs
Max size
per input base64-encoded image		 7 MB 7 MB 7 MB 7 MB 7 MB ---
Max number
of output images		 --- --- Up to output token limit Up to output token limit Up to output token limit ---
Video (per request)
Max number
of input video files		 10 files 10 files Up to input token limit Up to input token limit --- --- see docs
Max length
of all input video
(frames only)		 ~60 minutes ~60 minutes ~25 minutes ~12 minutes --- ---
Max length
of all input video
(frames+audio)		 ~45 minutes ~45 minutes --- --- --- ---
Audio (per request)
Max number
of input audio files		 1 file 1 file --- --- --- --- see docs
Max length
of all input audio		 ~8.4 hours ~8.4 hours --- --- --- ---

* For all Gemini models, a token is equivalent to about 4 characters, so 100 tokens are about 60-80 English words. For Gemini models, you can determine the total count of tokens in your requests using countTokens.

** For gemini-4-argon, Firebase AI Logic does not yet support long decode, but that feature is coming soon!

*** PDFs are treated as images, so a single page of a PDF is treated as one image. The number of pages allowed in a request is limited to the number of images the model can support.

Find additional detailed information


Model versioning and naming patterns

Models are offered in stable, preview, and experimental versions. For convenience, the -latest aliases are supported, but they're not recommended due to the inconsistency in which model version they point to.

Make sure to review our best practices for using model versions.

To find specific model names to use in your code, see the "available model names" section later on this page.

Version type /
Release stage		 Description Model name pattern
Stable Stable versions are available and supported for production use starting on the release date.
  • A stable model version is typically released with a retirement date, which indicates the last day that the model is available. After this date, the model is no longer accessible or supported by Google.
  • Most stable models are available for 12 months after their release date (even if a replacement model is released).
  • For the Agent Platform Gemini API (formerly Vertex AI), some stable models only have short-term availability, which means that they may be shut down (retired) as early as 45 days after a replacement model is released.

Model names of stable versions have no suffix

Examples:
gemini-4-argon
gemini-3.8-flash

Preview Preview versions have new capabilities and are considered not stable.
  • These models are not recommended for production use, come with more restrictive rate limits, and may have billing requirements.
  • These models are shut down (retired) within a few weeks or months after their associated stable version is released.
  • For the Agent Platform Gemini API (formerly Vertex AI), preview models are only available in the global location.

Model names of preview versions are appended with -preview

Example:
gemini-3-pro-preview

Experimental Experimental versions have new capabilities and are considered not stable.
  • These models are not recommended for production use and come with more restrictive rate limits. Experimental models are intended for gathering feedback and to enable experimentation with our latest features.
  • These models are shut down (retired) within a few weeks or months after their associated stable version is released.
  • For the Agent Platform Gemini API (formerly Vertex AI), experimental models are only available in the global location.

Model names of experimental versions are appended with -exp along with the model's release date (-MM-DD)

Example:
gemini-2.5-pro-exp-03-25
(released on March 25, 2025)

Shutdown (retired) Shutdown (retired) versions are past their shutdown (retirement) date and have been permanently deactivated.
  • Shutdown (retired) models are no longer accessible or supported by Google, and a request using a retired model name returns a 404 error.

---

Best practices for using model versions

  • In your production apps, use the explicit model name for the most recent stable version.

    For Agent Platform Gemini API (formerly Vertex AI), if you choose to use a short-term availability model in your production app, it's even more critical that you use Firebase Remote Config or server prompt templates to control the model name used for your AI feature.

  • Use preview and experimental versions only during prototyping. We recommend using a stable version when you start developing and testing for a production use case.

  • We do not recommend using the -latest aliases (even during development). This alias points to the latest release for a specific model variation (which could be a stable, preview, or experimental version). This alias will get hot-swapped with every new release of a specific model variation, and only for breaking changes will a 2-week-prior notification email be sent. This instability of which model you're actually using can lead to unexpected behavior changes for your AI feature.


Available model names

Model names are the explicit values that you include in your code during initialization of the model.

For initialization examples for your platform, see the getting started guide.

For details about the release stages (especially for use cases, billing, and shutdown), see model versioning and naming patterns.

Programmatically list all available models

You can list all available models names using the REST API:

Note that this returned list will include all models supported by the Gemini API providers, but Firebase AI Logic only supports the Gemini models described on this page.

General-use models

Model Gemini Developer API Agent Platform Gemini API
gemini-4-argon 1

Release date: October 13, 2026
Release stage: Stable 		Shutdown date: --- Shutdown date: ---
Model Gemini Developer API Agent Platform Gemini API
gemini-3.8-flash

Release date: September 2, 2026
Release stage: Stable 		Shutdown date: --- Shutdown date: ---
gemini-3.7-flash 2

Release date: August 13, 2026
Release stage: Stable 		Requests routed to gemini-3.8-flash
Replacement: gemini-3.8-flash 		Shutdown date: January 28, 2027
Replacement: gemini-3.8-flash
gemini-3.6-flash 2

Release date: July 21, 2026
Release stage: Stable 		Shutdown date: ---
Replacement: gemini-3.8-flash 		Shutdown date: November 19, 2026
Replacement: gemini-3.8-flash
gemini-3.5-flash

Release date: May 19, 2026
Release stage: Stable 		Requests routed to gemini-3.6-flash
Replacement: gemini-3.8-flash 		Shutdown date: ---
Replacement: gemini-3.8-flash
gemini-3.5-flash-lite

Release date: July 21, 2026
Release stage: Stable 		Shutdown date: --- Shutdown date: ---
gemini-3.1-flash-lite

Release date: May 7, 2026
Release stage: Stable 		Shutdown date: ---
Replacement: gemini-3.5-flash-lite 		Shutdown date: ---
Replacement: gemini-3.5-flash-lite
gemini-3.1-pro-preview

Release date: February 19, 2026
Release stage: Preview 		Shutdown date: ---
Replacement: gemini-4-argon 		Shutdown date: ---
Replacement: gemini-4-argon
Model Gemini Developer API Agent Platform Gemini API
gemini-2.5-pro

Release date: June 17, 2025
Release stage: Stable 		Requests routed to gemini-3.8-flash (starting December 1, 2026)
Replacement: gemini-3.8-flash 		Shutdown date: October 20, 2026
Replacement: gemini-3.8-flash
gemini-2.5-flash

Release date: June 17, 2025
Release stage: Stable 		Requests routed to gemini-3.5-flash-lite (starting December 1, 2026)
Replacement: gemini-3.5-flash-lite 		Shutdown date: October 20, 2026
Replacement: gemini-3.5-flash-lite
gemini-2.5-flash-lite

Release date: July 22, 2025
Release stage: Stable 		Shutdown date: ---
Replacement: gemini-3.5-flash-lite 		Shutdown date: October 20, 2026
Replacement: gemini-3.5-flash-lite

1 Requires the pay-as-you-go Blaze pricing plan regardless of your Gemini API provider.

2 For the Agent Platform Gemini API (formerly Vertex AI), this is a "short-term availability model". See Model versioning and naming patterns

Very old general-use models that are shutdown: gemini-1.0-pro, gemini-1.0-pro-001, gemini-1.0-pro-002, gemini-1.0-pro-vision, gemini-1.0-pro-vision-001 gemini-1.5-flash-001, gemini-1.5-flash, gemini-1.5-flash-002, gemini-1.5-pro-001, gemini-1.5-pro, gemini-1.5-pro-002, gemini-2.0-flash-lite, gemini-2.0-flash-lite-001, gemini-2.0-flash, gemini-2.0-flash-001

Image-generating (Nano Banana) models

Model Gemini Developer API Agent Platform Gemini API
gemini-nano-banana-2.1 1, 2

Release date: October 6, 2026
Release stage: Stable 		Shutdown date: --- Shutdown date: ---
gemini-3.1-flash-image 1

Release date: May 28, 2026
Release stage: Stable 		Shutdown date: --- Shutdown date: ---
gemini-3.1-flash-lite-image 1

Release date: June 30, 2026
Release stage: Stable 		Shutdown date: --- Shutdown date: ---
gemini-3-pro-image 1

Release date: May 28, 2026
Release stage: Stable 		Shutdown date: --- Shutdown date: ---

1 Requires the pay-as-you-go Blaze pricing plan regardless of your Gemini API provider.

2 For the Agent Platform Gemini API (formerly Vertex AI), this is a "short-term availability model". See Model versioning and naming patterns

Very old image-generating models that are shutdown: gemini-2.5-flash-image, imagen-3.0-fast-generate-001, imagen-3.0-generate-001, imagen-3.0-generate-002, imagen-3.0-capability-001, imagen-4.0-ultra-generate-001, imagen-4.0-fast-generate-001, imagen-4.0-generate-001

Audio - text-to-speech (TTS) models

Model Gemini Developer API Agent Platform Gemini API
gemini-3.1-flash-tts-preview

Release date: June 17, 2026
Release stage: Preview 		Shutdown date: --- Shutdown date: ---

Audio - Live API models

Model Gemini Developer API Agent Platform Gemini API
gemini-3.1-flash-live-preview

Release date: March 26, 2026
Release stage: Preview 		Shutdown date: --- Not supported
gemini-live-2.5-flash-native-audio

Release date: December 12, 2025
Release stage: Stable 		Not supported Shutdown date: ---
gemini-2.5-flash-native-audio-preview-12-2025

Release date: December 12, 2025
Release stage: Preview 		Shutdown date: ---
Replacement: gemini-3.1-flash-live-preview 		Not supported
gemini-live-2.5-flash-preview-native-audio-09-2025

Release date: September 18, 2025
Release stage: Preview 		Not supported Shutdown date: ---
Replacement: gemini-live-2.5-flash-native-audio
gemini-2.5-flash-native-audio-preview-09-2025

Release date: September 18, 2025
Release stage: Preview 		Shutdown date: ---
Replacement: gemini-3.1-flash-live-preview 		Not supported

Very old Live API models that are shutdown: gemini-2.0-flash-live-preview-04-09, gemini-2.0-flash-live-001, gemini-live-2.5-flash-preview


Supported languages

General-use models

All Gemini general-use models (like Gemini 4 Argon, Gemini 3.8 Flash, Gemini 3.5 Flash-Lite, and earlier 3.x models) can understand and respond in over 100 languages:

Afrikaans (af), Albanian (sq), Amharic (am), Arabic (ar), Armenian (hy), Assamese (as), Azerbaijani (az), Basque (eu), Belarusian (be), Bengali (bn), Bosnian (bs), Bulgarian (bg), Catalan (ca), Cebuano (ceb), Chinese simplified and traditional (zh), Corsican (co), Croatian (hr), Czech (cs), Danish (da), Dhivehi (dv), Dutch (nl), English (en), Esperanto (eo), Estonian (et), Filipino (Tagalog) (fil), Finnish (fi), French (fr), Frisian (fy), Galician (gl), Georgian (ka), German (de), Greek (el), Gujarati (gu), Haitian Creole (ht), Hausa (ha), Hawaiian (haw), Hebrew (iw), Hindi (hi), Hmong (hmn), Hungarian (hu), Icelandic (is), Igbo (ig), Indonesian (id), Irish (ga), Italian (it), Japanese (ja), Javanese (jv), Kannada (kn), Kazakh (kk), Khmer (km), Korean (ko), Krio (kri), Kurdish (ku), Kyrgyz (ky), Lao (lo), Latin (la), Latvian (lv), Lithuanian (lt), Luxembourgish (lb), Macedonian (mk), Malagasy (mg), Malay (ms), Malayalam (ml), Maltese (mt), Maori (mi), Marathi (mr), Meiteilon (Manipuri) (mni-Mtei), Mongolian (mn), Myanmar (Burmese) (my), Nepali (ne), Norwegian (no), Nyanja (Chichewa) (ny), Odia (Oriya) (or), Pashto (ps), Persian (fa), Polish (pl), Portuguese (pt), Punjabi (pa), Romanian (ro), Russian (ru), Samoan (sm), Scots Gaelic (gd), Serbian (sr), Sesotho (st), Shona (sn), Sindhi (sd), Sinhala (Sinhalese) (si), Slovak (sk), Slovenian (sl), Somali (so), Spanish (es), Sundanese (su), Swahili (sw), Swedish (sv), Tajik (tg), Tamil (ta), Telugu (te), Thai (th), Turkish (tr), Ukrainian (uk), Urdu (ur), Uyghur (ug), Uzbek (uz), Vietnamese (vi), Welsh (cy), Xhosa (xh), Yiddish (yi), Yoruba (yo), and Zulu (zu).

Image-generating models (Nano Banana)

The Gemini Nano Banana models can accept text prompts across many languages. However, they are optimized for best performance (both for text prompts and for rendering text within generated images) in the following languages and locales:

Arabic (ar-EG), German (de-DE), English (EN), Spanish (es-MX), French (fr-FR), Hindi (hi-IN), Indonesian (id-ID), Italian (it-IT), Japanese (ja-JP), Korean (ko-KR), Portuguese (pt-BR), Russian (ru-RU), Ukrainian (uk-UA), Vietnamese (vi-VN), and Simplified Chinese (zh-CN).

To learn more about prompt best practices and generating text in images, see Supported languages for image generation.

Audio-generating models (TTS and Live API)

Audio-generating models support specialized language sets optimized for speech synthesis and (for the Live API) real-time bidirectional dialogue:

  • Text-to-speech (TTS) models: The Gemini 3.x TTS models can automatically detect and generate speech in nearly 80 languages and dialects.

    To view the full list of supported languages and available BCP-47 locale codes, see Supported languages for speech generation.

  • Live API models: Live API models support real-time, bidirectional voice interactions in nearly 80 languages and dialects.

    To view the full list of supported languages and learn how to influence the spoken response language, see Supported languages for the Live API.


Information about previous models

This section provides information about previous generation models. For more information about the deprecation and shutdown of these older models (as well as recommended replacement models), see deprecated and shutdown models FAQ.

  • General-use models that are shutdown: gemini-1.0-pro, gemini-1.0-pro-001, gemini-1.0-pro-002, gemini-1.0-pro-vision, gemini-1.0-pro-vision-001 gemini-1.5-flash-001, gemini-1.5-flash, gemini-1.5-flash-002, gemini-1.5-pro-001, gemini-1.5-pro, gemini-1.5-pro-002, gemini-2.0-flash-lite, gemini-2.0-flash-lite-001, gemini-2.0-flash, gemini-2.0-flash-001

  • Image-generating models that are shutdown: gemini-2.5-flash-image, imagen-3.0-fast-generate-001, imagen-3.0-generate-001, imagen-3.0-generate-002, imagen-3.0-capability-001, imagen-4.0-ultra-generate-001, imagen-4.0-fast-generate-001, imagen-4.0-generate-001

  • Live API models that are shutdown: gemini-2.0-flash-live-preview-04-09, gemini-2.0-flash-live-001, gemini-live-2.5-flash-preview

If you can't find the information you're looking for in the following sub-sections, you can find even more information in your chosen API provider documentation: Gemini Developer API or Agent Platform Gemini API (formerly Vertex AI)

View details about about previous models


Next steps

Try out the capabilities of the Gemini API