Gemini
Gemini 3.6 Flash & 3.1 Pro — Google
Google's multimodal workhorse. Upload literally anything — text, images, audio, video, entire PDFs — and it'll process it all in a single prompt. Gemini 3.6 Flash is the newest release and burns up to 17% fewer tokens than 3.5 Flash; 3.1 Pro is still the top Pro tier you can actually run. Nano Banana keeps Gemini the best image generator of the big four.
3.6 Flash Released
Jul 21, 2026
Context
1M tokens
Starting Price
$7.99/mo (AI Plus)
Output
~66K tokens
01
Overview
Gemini is Google DeepMind's flagship AI, and right now its lineup is genuinely confusing. Gemini 3.6 Flash (July 21, 2026) is the newest and best-value model — a workhorse that improves on coding, knowledge work, and multimodal tasks while using up to 17% fewer tokens than 3.5 Flash. But the Pro tier hasn't moved since February: Gemini 3.1 Pro is still the most capable Gemini you can actually run.
Gemini 3.5 Pro has been "coming soon" since Google I/O in May. Google promised June, then July 17, then said it was testing with partners and would "land soon." Reporting points to internal delays around coding, math, and hallucination reliability. Until it ships, plan around 3.6 Flash for volume and 3.1 Pro for depth — not around a Pro model that keeps slipping.
What truly sets Gemini apart is its native multimodal architecture. While other models bolt on vision or audio as afterthoughts, Gemini was built from the ground up to process text, images, audio (up to 8.4 hours), video (up to 1 hour), and PDFs (up to 900 pages) all in a single prompt. No preprocessing, no separate APIs, no workarounds.
The Deep Research mode is a standout: it analyzes hundreds of sources in real-time and generates comprehensive research reports in minutes — think graduate-level literature reviews on demand.
Access for free at gemini.google.com, or start with Google AI Plus at $7.99/mo for expanded access. Power users will want AI Pro ($19.99/mo) for the full Gemini 3.1 Pro experience with Deep Research. If you need the heaviest reasoning firepower, AI Ultra ($249.99/mo) unlocks Deep Think mode and cutting-edge features. Google also shipped 3.5 Flash-Lite for high-volume work and 3.5 Flash Cyber for security — though Cyber is gated to governments and trusted partners in a limited pilot.
02
Key Capabilities
Natively Multimodal
Process text, images (up to 900/prompt), audio (8.4 hours), video (1 hour), and PDFs (900 pages) in one prompt
1M Token Context
1M-token window on both 3.6 Flash and 3.1 Pro, with roughly 66K max output
Nano Banana (2 / Pro / Lite)
Nano Banana 2 (Gemini 3.1 Flash Image) is the mainline image model; Nano Banana Pro (Gemini 3 Pro Image) adds 4K output and the best text rendering anywhere; a Lite variant shipped June 30 for volume work
Deep Research
Analyzes hundreds of sources in real-time for comprehensive research reports
Token Efficiency (3.6 Flash)
Up to 17% fewer tokens than 3.5 Flash for the same work — a real cost cut, not a marketing one, and it compounds on long agent runs
Deep Think Mode
Gemini 3.1 Deep Think for complex, multi-step problems — takes longer but nails hard reasoning tasks. Available to AI Ultra subscribers
Flash Cyber (Restricted)
Gemini 3.5 Flash Cyber is fine-tuned to find and fix security vulnerabilities — but it's limited to governments and trusted partners in a pilot program, not general access
SVG/3D Code Generation
Generates, animates, and renders SVG graphics and 3D code from natural language
Google Workspace Integration
Works across Gmail, Docs, Sheets, and other Google apps
77.1% on ARC-AGI-2
Strong performance on the reasoning benchmark
03
Pricing
| Tier | Price | Details |
|---|---|---|
| Free | $0 | Basic Gemini features, limited usage. |
| Google AI Plus | $7.99/mo | Expanded access. 50% off first 2 months. |
| Google AI Pro | $19.99/mo ($99.99/yr promo) | Gemini 3.1 Pro, Deep Research, 2TB storage, Workspace AI. Free first month. |
| Google AI Ultra | $249.99/mo | Highest limits, Deep Think mode, Veo 2/3 video gen, cutting-edge features. 50% off first 3 months. |
| API — 3.6 Flash | $1.50 / $7.50 per 1M tokens | The value pick. Cached input at $0.15/M. 1M context, ~66K output. |
| API — 3.1 Pro (≤200K) | $2 / $12 per 1M tokens | Still the top Pro tier. Input / output pricing. |
| API — 3.1 Pro (>200K) | $4 / $18 per 1M tokens | Extended context pricing. |
| API — 3.5 Flash-Lite | Cheapest in class | Built for developers running AI at scale where cost dominates. |
04
Best Use Cases
Processing large multimedia datasets (video, audio, images combined)
Research tasks requiring synthesis of hundreds of sources (Deep Research)
Google Workspace power users wanting AI deeply embedded in their workflow
Creative and design tasks involving SVG generation and 3D code
Cost-sensitive applications via the Flash-Lite tier
Large document and PDF analysis and extraction
Image generation for marketing, social media, and creative projects
05
Pro Prompts
Multimodal Analysis
Analyze this uploaded chart and forecast trends for the next two quarters. Flag assumptions you're least confident about.Deep Research
Use Deep Research to compile a comprehensive analysis of [industry/topic]. Include market size, key players, emerging trends, and risk factors. Cite every source.Image Generation
Generate a professional hero image for a tech startup landing page: dark theme, abstract geometric patterns, cyan accent lighting, minimal and modern. 16:9 aspect ratio.Video Analysis
Watch this 30-minute product demo video and create a structured summary: key features shown, claims made, competitor comparisons mentioned, and any technical specifications stated.06
Limitations
Long-context reliability degrades past ~120-150K tokens with rising "summary drift" and hallucinated content
Verbose outputs — generates 4x more tokens than average in evaluations, driving up costs
Structured output inconsistency — only 84% schema-valid JSON responses without retries
Strict rate limits — even Tier 1 paid users limited to 250 requests/day
Scanned PDFs need OCR preprocessing; raw uploads lead to mangled tables
Flash Cyber, the security-tuned model, is locked to governments and trusted partners — you can't just sign up for it
Gemini 3.5 Pro has slipped repeatedly since Google I/O in May — the Pro tier hasn't been updated since February, while rivals shipped new flagships in July
07
Multimodal Capabilities
Input
- Text
- Images (up to 900/prompt)
- Audio (up to 8.4 hours)
- Video (up to 1 hour)
- PDFs
- Code repositories
Output
- Text
- Code
- SVG graphics
- 3D code rendering
Image Generation
Via Nano Banana 2 (Gemini 3.1 Flash Image) by default, Nano Banana Pro (Gemini 3 Pro Image) for 4K and typography, or Nano Banana 2 Lite for volume. In the Gemini app: "Create images" from the tools menu, then Fast / Thinking / Pro
Video Generation
Veo 2/3 available on Ultra tier
08
Recent Updates
Jul 21, 2026
Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber released — 3.6 Flash cuts token usage up to 17% vs 3.5 Flash and improves coding, knowledge work, and multimodal. Still no 3.5 Pro.
Jun 30, 2026
Nano Banana 2 Lite announced — Google's fastest and most cost-efficient Gemini image model
Still Pending
Gemini 3.5 Pro announced at Google I/O (May 19, 2026) — reportedly 2M token context and an upgraded Deep Think mode. Promised for June, then July 17, still unshipped in August. Google says it's testing with partners.
Mar 2026
Gemini 3.1 Deep Think mode rolled out for AI Ultra subscribers — extended reasoning for complex problems
Mar 2026
Gemini 3.1 Flash Live launched — real-time voice and audio dialogue with low-latency streaming
Feb 26, 2026
Nano Banana 2 launched (built on Gemini 3.1 Flash Image) as default image generation across Google products — major upgrade in instruction following and text rendering
Feb 19, 2026
Gemini 3.1 Pro released with 2x reasoning improvement
Jan 2026
Google AI Plus plan ($7.99/mo) rolled out to all markets including the US