Gemini (Google)
Overview
The Gemini series is Google DeepMind's family of multimodal foundation models. They natively handle text, images, audio, video, and code, and they are deeply integrated with Google Workspace and Vertex AI. As of September 2026 the shape of the lineup is: Gemini 3.1 Pro as the current Pro tier, Gemini 3.8 Flash (released September 2, 2026, succeeding 3.7 Flash) as the newest fast tier, and Gemini 3 Deep Think for the hardest research and engineering work.
Current lineup
| Model | Best for | API pricing (input / output per 1M tokens) |
|---|---|---|
| Gemini 3.1 Pro | Current Pro tier: long context, hard tasks | $2 / $12 ($4 / $18 above 200K tokens) |
| Gemini 3.8 Flash | Newest Flash (successor to 3.7 Flash): coding, agents, multi-step reasoning | $0.75 / $3.75 (introductory through Dec 31, 2026; $1.50 / $7.50 from Jan 1, 2027) |
| Gemini 3 Deep Think | Deep reasoning for science, research, engineering | Google AI Ultra subscribers plus early-access API users only |
Gemini 3.8 Flash (same price and context length as 3.7 Flash, with sharper coding reasoning) costs $0.75 per 1M input tokens for the rest of 2026, or $0.075 for cached input (a 90% discount) — but the rate doubles on January 1, 2027, so do not build long-term cost models on it.
Key capabilities
1. Natively multimodal
One model handles text, images, audio, video, and code. It is strong on temporal understanding in video and on the structure of PDFs and codebases.
2. 1M-token context
Gemini 3.1 Pro carries a 1M-token context window — several books, an entire codebase, or hours of video in one request.
3. Thinking mode
Gemini 3.1 Pro is a thinking model: it reasons internally before answering hard questions.
4. Deep Research
Runs repeated autonomous web searches and produces a comprehensive, multi-source report. Well suited to academic and market research.
5. Google Workspace integration
Built into Gmail, Docs, Sheets, and Slides for drafting, summarizing, replying, and building presentations without leaving the document.
6. Gemini Live
Real-time voice conversation, including sharing a camera feed so you can ask questions about what it sees.
7. Computer control built into Flash
Reading the screen and clicking, typing, and scrolling is baked into the low-latency Flash tier rather than the flagship Pro, so browser and desktop automation needs no separate agent harness.
Context-length pricing
⚠️ The Gemini Pro tier uses tiered pricing above 200,000 tokens — once a prompt crosses the threshold, the entire request bills at the higher rate, not just the excess. Watch this on long-context work.
Gemini app and subscriptions
Since April 1, 2026, Pro-tier models (including 3.1 Pro) have been removed from the free tier and are paid-only.
| Plan | Monthly | What you get |
|---|---|---|
| Google AI (free) | $0 | Flash tier only; no Pro models |
| Google AI Plus | $7.99 | Entry-level paid tier |
| Google AI Pro | $19.99 | Gemini 3.1 Pro, Deep Research, 2TB storage |
| Google AI Ultra (5x) | $99.99 | 5x the Pro allowance, plus Gemini 3 Deep Think |
| Google AI Ultra (20x) | $199.99 | 20x the Pro allowance |
| Vertex AI | Pay-per-use | Enterprise API, private deployment |
Typical uses
- Multimodal analysis: long documents that include video, audio, and images
- Deep Research: generating literature reviews and market reports
- Google Workspace work: Gmail replies, Docs drafts, Sheets analysis
- Codebase comprehension: Q&A and refactoring across a whole project
- Video summarization: chapters and summaries for long recordings
Strengths
- ✅ Natively multimodal (video and audio included)
- ✅ Google ecosystem integration (Search, Workspace, YouTube, Vertex AI)
- ✅ Gemini 3.8 Flash at $0.75 input through year-end is cheap for coding and agent work (and sharper on reasoning than 3.7 Flash)
- ✅ 1M-token context
- ✅ Enterprise-grade via Vertex AI
- ✅ Gemini Live — real-time voice and video conversation
Weaknesses
- ❌ Answer quality can be inconsistent
- ❌ Frequent API changes
- ❌ Tiered pricing above 200K tokens makes long contexts expensive
- ❌ The Gemini 3.8 Flash introductory rate expires at the end of 2026 and doubles in January 2027
- ❌ An ongoing DeepMind reorganization and senior-researcher departures leave the development org's continuity unclear
Official resources
- Website: https://deepmind.google
- Gemini app: https://gemini.google.com
- Google AI Studio: https://aistudio.google.com
- Vertex AI: https://cloud.google.com/vertex-ai