| Category | Details |
|---|---|
| Model Name | Gemini 3 |
| Released by | Google and Google DeepMind |
| Launch Date | November 18, 2025 |
| Top Features | Multimodal reasoning, agentic coding, 1M-token context, Deep Think mode |
| Key Innovations | 1501 Elo (LMArena), PhD-level reasoning, 87.6% (Video-MMMU), 72.1% (SimpleQA Verified), 93.8% (GPQA Diamond), 45.1% (ARC-AGI-2) |
| Use Cases | Learning (flashcards, cookbooks, training plans), coding (WebDev Arena leaderboard, Terminal-Bench 2.0), planning (Vending-Bench 2, schedules), autonomous task execution |
| New Platform | Google Antigravity: Agent-first developer platform for autonomous software development |
| Availability | Google Search (AI Mode), Gemini app (Pro and Ultra), Google AI Studio, Vertex AI, Gemini CLI, Google Antigravity, Gemini 3 Deep Think (post safety tests) |
| Safety Measures | Enhanced safeguards against prompt injection, biases, misuse; tested with UK AISI, Vaultis, Apollo, and Frontier Safety Framework |
| Developer Tools | Gemini 3 available on Google AI Studio, Gemini CLI, Google Antigravity, GitHub, Replit, JetBrains |
| Benchmarks Achieved | 1487 Elo (WebDev Arena), 54.2% (Terminal-Bench 2.0), 76.2% (SWE-bench Verified) |
Workflow
Login