Google’s AI ecosystem is a mess of overlapping brands. Google Labs, AI Studio, Flow, Antigravity, and Vibe Coding sound interchangeable but serve completely different purposes. Here’s the 2-minute breakdown so you stop clicking the wrong tool.


The TL;DR Matrix

ToolWhat It Actually IsBest ForCostProduction Ready?
Google LabsExperimental playgroundTrying bleeding-edge demosFree❌ No export
AI StudioPrototyping environmentTesting Gemini models, API keysFree tier + API⚠️ Manual export only
FlowCreative video studioGenerating video with Veo 3.1$20-250/mo❌ Labs-only
AntigravityAgentic IDE (like Cursor)Coding with autonomous agents$20-250/mo⚠️ Export to GitHub
Vibe CodingApp prototypingBuilding apps from natural languageFree tier + API⚠️ Code export available

The confusion: Google Labs contains Flow, Antigravity, and Vibe Coding as “experiments.” AI Studio is the gateway to production API access. They’re adjacent but serve different stages of the workflow.


Google Labs: The Playground

What it is: A sandbox for Google’s experimental AI projects. Everything here is labeled “experimental” which means it can break, change, or disappear without notice.

What’s inside:

  • Flow — Video generation with Veo 3.1
  • Antigravity — Agentic coding IDE
  • Vibe Coding — Natural language app builder
  • Whisk — Image blending and style transfer
  • GenTabs — Browser session cloning
  • Dozens of smaller experiments

The catch: No native export. Your Flow videos download as MP4s without metadata. Your Antigravity code pushes to GitHub but loses the agent context. Everything stays in the sandbox unless you manually extract it.

When to use: Early exploration, creative prototyping, testing concepts before committing engineering resources.


AI Studio: The Prototyping Workbench

What it is: The production on-ramp. Free access to Gemini models with rate limits, plus API key management for scaling to production.

Key features:

  • Free tier: 100-1000 requests/day on Gemini 3 Flash/Pro
  • API keys: Direct path to Vertex AI production deployment
  • Prompt playground: Test system prompts, temperature, token limits
  • Cost calculator: See exactly what production usage costs

Critical distinction: AI Studio is where you validate that a Gemini model can handle your use case before you commit to Google’s production infrastructure (Vertex AI).

Free access details: See Gemini 3 Flash free tier guide


Flow: Creative Video Generation

What it is: A creative storytelling tool powered by Veo 3.1. Generates video with native audio synchronization through Lyria 2 music generation.

Technical specs:

  • Context window: 1M-2M tokens (enables feature-length screenplays)
  • Video length: Up to 10 seconds per generation
  • Pricing: $0.02-0.50/second depending on quality tier
  • Audio: Native synchronization (unique differentiator vs Runway/Sora)

The “Ingredients” feature: Upload up to 3 reference images (character, scene, style) and Flow maintains consistency across video sequences. This solves the “character drift” problem that plagues other video generators.

The problem: No API export. You download MP4s and lose generation parameters. Can’t reproduce the same style systematically for production workflows.

Competitor comparison: Runway Gen-4 has clearer pricing and production integration. Sora 2 is research preview only. Flow wins on context size and audio sync but loses on deployment readiness.


Antigravity: The Agentic IDE

What it is: Google’s answer to Cursor and Windsurf. An agentic coding environment built on Gemini 3 Pro with 1M token context.

Core capabilities:

  • Agent-assisted mode: Human-in-the-loop for sensitive operations
  • Agent-driven mode: Full autonomy with configurable allow/deny lists
  • Multi-platform deployment: Chat, Slack, Discord, SMS integrations
  • 42-second median: Full-stack feature generation time (per Google’s benchmarks)

The controversy: Google bundled Claude Opus 4.5 as a key feature, then imposed severe rate limits. Pro subscribers ($20/mo) report 4-7 day lockouts after 2 hours of Claude use. The “5-hour refresh” advertised only applies to Gemini models.

Reality check: If you need Claude Opus access, pay Anthropic directly. If you want an agentic IDE, Cursor and Windsurf have clearer pricing and better production integration.

Benchmark performance:

  • SWE-bench Verified: 76.2% (vs Claude 4.5 Opus at 80.9%)
  • Context window: 1M tokens (5× larger than Claude 3.5)
  • Cost: $0.50-1.25/1M input tokens (2.5-6× cheaper than Claude)

When to choose: High-volume workflows where the 1M context window matters more than that final 4% of reasoning performance.


Vibe Coding: App Prototyping

What it is: Natural language application development inside AI Studio. Describe what you want (“customer dashboard with Stripe integration”) and get working React/Vue components.

How it differs from Antigravity:

  • Vibe Coding: Web app prototyping, exports to GitHub/ZIP
  • Antigravity: Full agentic IDE with terminal integration, multi-file refactoring

Use case: Product managers validating concepts before engineering handoff. Not for production systems—the generated code needs review and hardening.


The Production Gap

The structural problem: Labs tools prototype in isolation. Moving to production requires:

  1. Export code/assets manually
  2. Reconfigure for Vertex AI
  3. Set up CI/CD separately
  4. Implement observability from scratch

Estimated friction: 15-20 manual steps per deployment. Compare to Cursor (native VS Code extension) or Windsurf (integrated deployment).

Recommended architecture:

[Antigravity/Vibe Labs Prototype]
         ↓
[Manual Export: Code + Dependencies]
         ↓
[GitHub Repository]
         ↓
[Vertex AI Production Deployment]

Which One Should You Use?

Choose AI Studio when:

  • Testing Gemini models for production use
  • Need free tier access to Flash/Pro
  • Building API integrations

Choose Flow when:

  • Previsualizing creative concepts
  • Need audio-synced video generation
  • Context size matters (1M-2M tokens)

Choose Antigravity when:

  • High-volume coding workflows
  • 1M context window is critical
  • Cost efficiency trumps absolute performance

Skip Labs entirely when:

  • You need production deployment paths
  • Time-sensitive workflows (rate limit lockouts)
  • Require reproducible pipelines

Google models and pricing:

Competitor comparisons:

Free tier guides:


Last updated: February 9, 2026. Based on Google Labs Technical Audit (Feb 2026) and hands-on testing.