Gemini
Gemini 3.8 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 3.8 Flash is the newest workhorse at a cheap introductory price that doubles on Jan 1; 3.1 Pro is still the deepest Gemini you can actually run. Gemini 4 Argon is announced, but invite-only. Nano Banana keeps Gemini the best image generator of the big four.
3.8 Flash Released
Sep 2, 2026
Context
1M tokens
Starting Price
$4.99/mo (AI Plus)
Output
~66K tokens
01
Overview
Gemini is Google DeepMind's flagship AI, and right now its lineup is genuinely confusing. Gemini 3.8 Flash (GA September 2, 2026) is the newest workhorse — Google's most capable Flash, at an introductory $0.75 / $3.75 per million tokens. Read the fine print: that price runs through December 31, then doubles to $1.50 / $7.50. Meanwhile the Pro tier still hasn't moved since February: Gemini 3.1 Pro (still labelled Preview) is the deepest Gemini you can actually run.
Remember Gemini 3.5 Pro, "coming soon" since Google I/O in May? It never shipped. Google skipped it: on September 30 it announced Gemini 4 Argon instead, and Argon has effectively replaced it. The catch: Argon is invite-only through the Fairwind Program for "trusted cyber defenders," with no public model ID. Google says broader access comes later, starting with paid API customers and AI Ultra — no date. Announced pricing is $2 / $10 introductory, then $4 / $20. Until you can actually call it, plan around 3.8 Flash for volume and 3.1 Pro for depth — not around a model you can't use.
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 — the free app runs 3.6 Flash, with varying access to 3.1 Pro. Google AI Plus is $4.99/mo for expanded access. In the app, 3.8 Flash is AI Pro and Ultra only: AI Pro ($19.99/mo) gets you 3.8 Flash plus the full Gemini 3.1 Pro experience with Deep Research. If you need the heaviest reasoning firepower, AI Ultra ($99.99–$199.99/mo) unlocks Deep Think mode and is first in line for Argon when it opens up. Google also ships 3.5 Flash-Lite for high-volume work, and a security-tuned 3.8 Flash Cyber — though, like Argon, Cyber is locked to the Fairwind Program.
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
~1.05M-token window on 3.8 Flash and 1M on 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
Cheap by Default (3.8 Flash) — For Now
Introductory $0.75 / $3.75 per million through Dec 31, 2026, then $1.50 / $7.50. Great while it lasts — but budget for the doubled price, because it compounds on long agent runs too
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 3.8 Flash Cyber is the security-tuned Flash for finding and fixing vulnerabilities — but it's limited to the invite-only Fairwind 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 | Gemini 3.6 Flash, varying access to 3.1 Pro, limited usage. No 3.8 Flash. |
| Google AI Plus | $4.99/mo | Expanded access. 50% off first 2 months. |
| Google AI Pro | $19.99/mo ($99.99/yr promo) | Gemini 3.8 Flash, 3.1 Pro, Deep Research, 2TB storage, Workspace AI. Free first month. |
| Google AI Ultra | $99.99–$199.99/mo | Highest limits, 3.8 Flash, Deep Think mode, Veo video gen. Named for early Argon access — no date yet. 50% off first 3 months. |
| API — 3.8 Flash | $0.75 / $3.75 per 1M tokens | The value pick — but introductory only, through Dec 31 2026. Jumps to $1.50 / $7.50 after that. ~1.05M context, ~66K output. |
| API — 3.1 Pro (≤200K) | $2 / $12 per 1M tokens | Still the deepest generally available Gemini (labelled Preview). Argon's announced $2 / $10 intro price undercuts it on output, but you can't buy Argon yet. |
| 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, high-volume work via 3.8 Flash (while the intro price lasts) or 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
3.8 Flash's cheap price is a promo: it doubles to $1.50 / $7.50 on Jan 1, 2027. Flash Cyber, the security-tuned model, is locked to the Fairwind Program — you can't just sign up for it
No new Pro you can actually use since February — 3.5 Pro never shipped, and Gemini 4 Argon, which has effectively replaced it, is invite-only with no public model ID and no date for wider access
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
Sep 30, 2026
Gemini 4 Argon announced — Google skips 3.5 Pro entirely. Invite-only via the Fairwind Program for trusted cyber defenders, no public model ID. Broader access "later," starting with paid API customers and AI Ultra; no date. Announced at $2 / $10 introductory, then $4 / $20.
Sep 22, 2026
Gemini 3.8 Flash TTS released for text-to-speech
Sep 15, 2026
Gemini 3.8 Live launched for real-time voice conversations
Sep 3, 2026
Lyria 3.5 music generation released
Sep 2, 2026
Gemini 3.8 Flash went GA — Google's most capable Flash, three weeks after 3.7. Same introductory $0.75 / $3.75 through Dec 31, then $1.50 / $7.50. In the Gemini app it's AI Pro and Ultra only; the free tier stays on 3.6 Flash.
Aug 13, 2026
Gemini 3.7 Flash went GA — Google's "most capable Flash model", ~1.05M context, and introductory pricing at about half of 3.6 Flash. Three weeks after 3.6. Still no 3.5 Pro.
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
May 19, 2026
Gemini 3.5 Pro announced at Google I/O — reportedly 2M token context and an upgraded Deep Think mode. Promised for June, then July 17. It never shipped; Gemini 4 Argon has effectively replaced it.
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