🇺🇸 USA · Google Gemini models lineup
Status: 🟩 COMPLETE 🟦 LIVING
Last updated: 2026-06-26
Plain-English tagline: Google DeepMind’s Gemini family — Pro, Flash, Nano variants; Deep Think reasoning mode; plus sibling models Imagen, Veo, Lyria. The full Google AI catalog explained.
Front-matter facts
Field Value Vendor Google DeepMind (Mountain View + London) Country / origin 🇺🇸 USA + 🇬🇧 UK Recommended for Australian users? ✅ Yes Privacy summary Per-product privacy; Vertex AI no-train; free AI Studio trains by default Free tier Generous via Gemini consumer app, Google AI Studio (with training caveat), AI Pro / Ultra Paid tiers Google AI Pro, Ultra, Vertex AI, Workspace tiers First released Gemini 1.0 December 2023; Gemini 3 family 2026 Last reviewed 2026-06-26 Official site https://deepmind.google/technologies/gemini
What it is
Google’s Gemini family across capabilities.
Chat / general-purpose
Model Best for Context window Gemini 3 Pro Default best-quality 1M tokens Gemini 3 Flash Cheap / fast everyday use 1M tokens Gemini 3 Flash-Lite Cheapest tier 1M tokens Gemini 3 Nano On-device (Pixel, Android) Smaller Gemini 2.5 Pro / Flash (legacy)Still available Same Gemini 1.5 Pro / Flash (legacy)Older, deprecated for new Same
Extended reasoning
Model Best for Gemini 3 Deep Think Hardest reasoning; extended thinking mode
Image generation
Model Best for Imagen 4 Current flagship; best photorealism Imagen 3 (legacy)Replaced by Imagen 4
Video generation
Model Best for Veo 3 Current flagship; with audio Veo 2 (legacy)Replaced
Music
Model Best for Lyria 2 Text-to-music gen
Voice / speech
Model Best for Gemini Live Real-time voice conversation Cloud Speech-to-Text Transcription Cloud Text-to-Speech Voice generation
Open-weight (Google’s open releases)
Model Best for Gemma 3 familyOpen-weight versions (smaller than Gemini Pro); Apache-friendly license Gemma Code Coding variant Gemma Multimodal Vision variant
Embeddings
Model Best for text-embedding-005 (or similar current)RAG embeddings via Vertex AI
Specialised research (DeepMind)
Model Best for AlphaFold 3 Protein structure (Nobel Prize 2024) AlphaProof + AlphaGeometry Mathematical proofs AlphaEarth Earth-system modelling
How to choose
Question Answer Best Gemini quality, broad use? Gemini 3 Pro Cheap / fast at scale? Gemini 3 Flash Cheapest tier? Gemini 3 Flash-Lite Hard reasoning? Gemini 3 Deep Think On-device (mobile)? Gemini 3 Nano (Pixel only)Image generation? Imagen 4 Video generation? Veo 3 Music? Lyria 2 Real-time voice? Gemini Live Open-weight to self-host? Gemma 3 Research-grade scientific AI? AlphaFold / AlphaProof / AlphaEarth
Access paths
Gemini consumer chat : gemini.google.com (free + AI Pro + Ultra)Google AI Studio : developer playground (free with training caveat)Vertex AI : production API with no-train + AUS data residencyWorkspace : Gemini in Gmail / Docs / Sheets / Slides / MeetPixel devices : Gemini Nano on-deviceNotebookLM, Antigravity, Whisk, Flow : Google product surfaces using Gemini
Recent changes
2026: Gemini 3 family; Imagen 4; Veo 3; Lyria 2; Deep Think2025: Gemini 2.5 series; AlphaProof + AlphaGeometry IMO 20242024: Gemini 2.0 family; Project Astra preview; Gemma 22023: Gemini 1.0 launch (rebranded from Bard)
Gotchas
Naming continuity — Gemini 3 succeeded 2.5 succeeded 2.0 succeeded 1.5 succeeded 1.0Free AI Studio trains on inputs — important privacy caveatGemma is the open-weight sibling family — Apache-friendly license; alternative to Llama / Mistral open weightsDeepMind = UK (London) heritage — Google acquired 2014; significant research work still in LondonPixel + Gemini Nano integration — for Aussie Pixel users, on-device AI is real
See also
Sources