Gemini

Google Gemini

Google

Google's Gemini 3.1 Pro delivers a 1M token context window with native multimodal understanding across text, image, video, and audio. The Flash-Lite variants offer some of the cheapest inference in the market.

gemini.google.com
Last updated: April 20, 2026
$2/M
3.1 Pro Input
1M
Context Window
66K
Max Output
4
Modalities
★★★★★
Coding
★★★★★
Reasoning
★★★★
Writing
★★★★★
Speed

Available Models

ModelInput $/1MOutput $/1MContextBest For
Gemini 3.1 Pro (flagship)$2.00$12.001MComplex reasoning, agentic coding
Gemini 3 Flash$0.40$1.601MBalanced speed + capability
Gemini 3.1 Flash-Lite$0.15$0.601MCost-efficient multimodal
Gemini 2.5 Pro$1.25$10.001MPrevious gen, stable
Gemini 2.5 Flash$0.15$0.601MBudget reasoning

Note: Pricing doubles for inputs exceeding 200K tokens on Vertex AI (Vertex AI Pricing).

Consumer Plans

PlanPriceIncludes
Free$0/moBasic Gemini access
Google AI Plus$8/moEnhanced Gemini, priority
Google AI Pro$20/moGemini 3 Pro, 2TB storage

Strengths & Weaknesses

Strengths
  • True multimodal: text, image, video, and audio natively (Google AI Docs)
  • 1M token context for long documents and codebases
  • Strong agentic workflows and autonomous coding
  • Controllable thinking budgets for cost management
  • Flash-Lite variants among cheapest in market
Weaknesses
  • Higher pricing for extended context (>200K tokens doubles cost)
  • Preview status may have rate limits and deprecation risks
  • 22-40% on some multi-hop noisy web reasoning benchmarks

Best For

  • Multimodal apps processing text, images, video, and audio
  • Long-context analysis of documents and codebases
  • Agentic workflows and coding agents
  • Enterprise AI via Vertex AI integration
  • Cost-efficient high-volume tasks (Flash-Lite)

Latest Release NEW

Gemini 3.1 Pro Preview — March 2026

  • Improved agentic and coding capabilities over Gemini 3.0
  • 1M token context window with multimodal input support
  • Enhanced tool use and function calling for complex workflows
  • Gemini 3 Pro Preview shut down March 9, 2026
  • Gemini 2.0 Flash scheduled for deprecation June 2026

Previous Releases

Gemini 2.0 Flash — December 2024

  • Speed-optimized model with native tool use and code execution
  • Multimodal inputs including image, video, and audio
  • Scheduled for deprecation June 2026 — migrate to 3.x

Gemini 1.5 Pro — February 2024

  • Pioneered the 1M token context window for commercial models
  • First model to achieve near-perfect recall on long-context benchmarks
  • Legacy model — still available but superseded by 3.x family