Gemini 3 Flash: Google's Fast New AI Model is Here
Google just released Gemini 3 Flash, a faster and cheaper version of Gemini 3 Pro. It's now the default model in the Gemini app worldwide.
Google just released Gemini 3 Flash, a faster and cheaper version of Gemini 3 Pro. It's now the default model in the Gemini app worldwide.
What is Gemini 3 Flash?
Gemini 3 Flash delivers "frontier intelligence built for speed at a fraction of the cost." It matches Gemini 3 Pro performance in many benchmarks while being 3x faster and significantly cheaper.
Known Codenames
Before release, Google tested Gemini 3 Flash on LM Arena under these codenames:
- Lithiumflow — efficiency-optimized variant
- Ghost Falcon — final testing version
- Oceanstone — earlier test version
- Fierce Falcon — Pro GA variant
Key Specs
| Feature | Gemini 3 Flash |
|---|---|
| Release Date | December 17, 2025 |
| Speed | 3x faster than 2.5 Pro |
| Default Model | Yes (Gemini App globally) |
| Best For | Video analysis, data extraction, visual Q&A |
Pricing
| Tier | Price |
|---|---|
| Input | $0.50 per million tokens |
| Output | $3.00 per million tokens |
Slightly higher than Gemini 2.5 Flash ($0.30/$2.50) but with significantly better performance.
Benchmarks
| Benchmark | Gemini 3 Flash | Gemini 3 Pro |
|---|---|---|
| MMMU-Pro | 81.2% | 81% |
| Humanity's Last Exam | 33.7% | 37.5% |
| GPQA Diamond | 90.4% | 91.9% |
Gemini 3 Flash even beats Pro on MMMU-Pro, Toolathlon, and MPC Atlas benchmarks.
Who's Using It
Major companies already on board:
- JetBrains
- Figma
- Cursor
- Harvey
- Latitude
How to Access
Consumers:
- Gemini App (default model)
- AI Mode in Google Search
Developers:
- Google AI Studio (preview)
- Vertex AI
- Gemini CLI
- Google Antigravity
- Android Studio
Model Picker Options
In Gemini App, you'll see:
- "Fast" — Gemini 3 Flash for quick answers
- "Thinking" — Gemini 3 Flash for complex problems
- "Pro" — Gemini 3 Pro for advanced math/code
Gemini 3 Flash is Google's new workhorse model — fast, affordable, and powerful enough for most tasks. Perfect for bulk processing, real-time apps, and cost-conscious developers.
