Gemini 3
#7 v Multimodální modelygoogle · v3 · od 2025-11-18 (Gemini 3 Pro Preview); Gemini 3 Flash: 2025-12-17 · 129× · naposledy 30. 6. 2026
73
Momentum
Gemini 3 is the third generation of Google DeepMind's multimodal model family, comprising Gemini 3 Pro (flagship) and Gemini 3 Flash (efficiency tier). The family is optimized for complex reasoning, multimodal understanding, and agentic workflows, defaulting to dynamic thinking. Access is available via the Gemini API (AI Studio), Vertex AI, the Gemini app, and third-party platforms; all Gemini 3-generation models are currently in preview status under a proprietary license.
Vývoj momenta
04.04.03.07.
Vlastnosti
| Key Benchmark (%) | Gemini 3 Pro: GPQA Diamond 91.9%, ARC-AGI-2 31.1% (45.1% Deep Think), AIME 2025 95% (no tools), SWE-Bench Verified 76.2%, MMMU-Pro 81.0% |
| Context Window (Tokens) | 1M token input, up to 64,000 token output (Gemini 3 Pro & Flash) |
| License | Proprietary (no open weights; preview status; Pre-GA Offerings Terms from Google Cloud) |
| Multimodality | Text, images, audio, video, PDFs, code repositories as input; no image segmentation support in Gemini 3 Pro/Flash |
| Platform | Gemini API (AI Studio), Vertex AI, Gemini App, Google Antigravity, Gemini CLI; third-party: Cursor, GitHub, JetBrains, Replit, and others |
| Price | API: Gemini 3 Flash $0.50/$3.00 per 1M tokens; Gemini 3.1 Pro $2.00/$12.00 (≤200K). Consumer: free (Flash), AI Pro $19.99/month, AI Ultra $99.99–$199.99/month |
| Price per 1M Tokens | Gemini 3 Flash: $0.50 input / $3.00 output; Gemini 3 Pro (deprecated, was preview): possibly higher – current successor Gemini 3.1 Pro: $2.00/$12.00 (≤200K tokens) |
| Release Date | Gemini 3 Pro: Nov 18, 2025; Gemini 3 Flash: Dec 17, 2025 (both preview) |