Model · Google · Reasoning
Gemini 2.5 Deep Think
Google's enhanced reasoning mode on top of Gemini 2.5 Pro. Trades latency for depth on hard math, science, and multi-step problem-solving.
- Modality
- Reasoning
- License
- Proprietary (Proprietary)
- Context window
- 1,000,000 tokens
- Released
- August 5, 2025
- Last verified
- May 10, 2026
- Runs locally
- No
Strengths
- 1M-token context window persists into reasoning mode
- Strong on math olympiad and scientific reasoning
- Multimodal inputs (text, image, video) carry into reasoning
Weaknesses
- Latency is significant — minutes per response on hard prompts
- Closed weights
- Available only on higher-tier Gemini plans
Try it
| Where | Type | Notes |
|---|---|---|
| Gemini app (Pro/Ultra plans) | product | Higher-tier plan required |
| Google AI Studio | hosted-api | API key required |
Version history
- Gemini 2.5 Deep Think Aug 2025 Current
- Gemini 2.5 Flash Apr 2025
- Gemini 2.5 Pro Mar 2025
Official sources
- Model docs docs
Change log
- — Initial entry.
Esc