한국어EN
Google

The Gemini family

Google's flagship AI model family, shown with the earlier T5 and specialist models for images, robots and more.

Only versions we have a record for. Missing ones appear as records are added.

21versions2018–2026Latest · Gemini 4 Argon

At a glance

The Gemini family — 21 versionsEach Gemini version from 2018 to 2026 as a dot, in release order. The list below has the same information.Before GeminiBERT · Oct 11, 2018T5 · Oct 23, 2019LaMDA · May 18, 2021PaLM · Apr 4, 2022Bard · Feb 6, 2023PaLM 2 · May 10, 2023Main lineGemini 1.0 · Dec 6, 2023Gemini 1.5 Pro · Feb 15, 2024Gemini 2.0 Flash · Dec 11, 2024Gemini 2.0 Flash Thinking Experimental · Dec 19, 2024Gemini 2.5 Pro Experimental · Mar 25, 2025Gemini 3 Pro · Nov 18, 2025Gemini 3.8 Flash / 3.8 Flash Cyber · Sep 2, 2026Gemini 4 Argon · Sep 30, 2026Specialist models(images, robots, mo…Gemini Robotics / Robotics-ER · Mar 12, 2025Gemini 2.5 Flash Image (nano-banana) · Aug 26, 2025Gemini Robotics 1.5 / ER 1.5 · Sep 25, 2025Gemini 2.5 Computer Use · Oct 7, 2025Nano Banana Pro (Gemini 3 Pro Image) · Nov 20, 2025Gemini Robotics 2 / ER 2 · Jul 30, 2026Gemini 3.8 Live / 3.8 Live Extended Thinking · Sep 15, 2026
Where labels would overlap, only the dot is drawn — point at it to see the name; every version is in the list below.

21 versions

In release order. Each date comes from the record that documents it — open a name to see the record.

  1. BERT

    Introduced encoder pre-training that reads context in both directions

    Before GeminiTextOpen weights

  2. T5

    Paper released together with the C4 dataset

    Before GeminiText

  3. LaMDA

    Announced as a model trained for dialogue

    Before GeminiText

  4. PaLM

    Announced as a 540B-parameter model

    Before GeminiText

  5. Bard

    Announced as a conversational service built on LaMDA

    Before GeminiText

  6. PaLM 2

    Came in four sizes, trained on over 100 languages

    Before GeminiText

  7. Gemini 1.0

    Natively multimodal: all input types trained together

    Main lineMultimodal

  8. Gemini 1.5 Pro

    128K context standard, 1 million tokens in preview

    Main lineMultimodal

  9. Gemini 2.0 Flash

    First 2.0 model, with native image and audio output

    Main lineMultimodal

  10. Gemini 2.0 Flash Thinking Experimental

    Reasoning model that shows its thinking to users

    Main lineReasoning

  11. Gemini Robotics / Robotics-ER

    Built on Gemini 2.0 to link knowledge to robot control

    Specialist models (images, robots, more)Multimodal

  12. Gemini 2.5 Pro Experimental

    Signalled that thinking would be built into all models

    Main lineReasoning

  13. Gemini 2.5 Flash Image (nano-banana)

    Image model for character-consistent edits, with SynthID

    Specialist models (images, robots, more)Multimodal

  14. Gemini Robotics 1.5 / ER 1.5

    Thinks before acting; skills transfer across robot types

    Specialist models (images, robots, more)Multimodal

  15. Gemini 2.5 Computer Use

    A dedicated model specialized for operating browsers

    Specialist models (images, robots, more)Multimodal

  16. Gemini 3 Pro

    Natively multimodal; strong at video and documents

    Main lineMultimodal

  17. Nano Banana Pro (Gemini 3 Pro Image)

    Draws accurate infographics and text in many languages

    Specialist models (images, robots, more)Multimodal

  18. Gemini Robotics 2 / ER 2

    Whole-body robot control; ER 2 opened in AI Studio

    Specialist models (images, robots, more)Multimodal

  19. Gemini 3.8 Flash / 3.8 Flash Cyber

    Shipped with Cyber, a security-focused variant

    Main lineReasoning

  20. Gemini 3.8 Live / 3.8 Live Extended Thinking

    Real-time voice models in two thinking-depth tiers

    Specialist models (images, robots, more)Multimodal

  21. Gemini 4 Argon

    Model for long expert work, output limit raised to 1M tokens

    LatestMain lineReasoning