Vercel adds Gemini 3.6 Flash to AI Gateway

Key Takeaways

• Gemini 3.6 Flash produces cleaner web and app development output while consuming fewer tokens
• Gemini 3.5 Flash-Lite targets subagent workflows for scoped tasks within larger agentic systems
• AI Gateway charges no platform fee on inference, including bring-your-own-key requests

Vercel has added Google's Gemini 3.6 Flash and Gemini 3.5 Flash-Lite to its AI Gateway, giving developers two more options for building AI-powered applications through a unified API. The new models focus on efficiency: fewer tokens consumed, fewer model calls made, and cleaner output for web development tasks.

The timing is notable. Google recently shipped three new Gemini models, and Vercel's quick integration signals that AI Gateway is keeping pace with provider releases. For engineering teams already using the platform, switching to these models requires changing a single string in their codebase.

What do these models actually do better?

Gemini 3.6 Flash improves on its predecessor across coding, agentic tasks, and web development. The practical claim: it produces cleaner app and web output while burning through fewer tokens per request. For teams paying per token, that efficiency translates directly to lower bills.

Gemini 3.5 Flash-Lite serves a different purpose. It upgrades the agentic capabilities of the Flash-Lite tier, positioning it as a fit for subagents. Think of it as the worker model you assign to handle scoped parts of a larger task. A parent agent orchestrates, Flash-Lite executes the bounded subtasks.

The agentic framing matters. Modern AI applications increasingly rely on multi-model orchestration, using cheaper, faster models for routine work while reserving expensive models for complex reasoning. Flash-Lite slots into that architecture.

How AI Gateway handles the integration

Using the new models through Vercel's AI SDK is straightforward. Set the model parameter to google/gemini-3.6-flash or google/gemini-3.5-flash-lite and call streamText. The SDK handles the rest.

import { streamText } from 'ai';

const result = streamText({
  model: 'google/gemini-3.6-flash',
  prompt: 'Build a settings page with profile and notification sections.',
});

AI Gateway provides more than a pass-through to Google's API. It bundles usage tracking, cost monitoring, automatic retries, failover routing, and performance optimizations. The pitch: higher uptime than going direct to providers.

Zero Data Retention support is included, which matters for teams building in regulated industries. Budget controls let you cap API key spending before a runaway loop drains your account. Custom reporting surfaces which models your team actually uses and what they cost.

The pricing model

Vercel makes an explicit claim here: AI Gateway reflects provider pricing with no markup and charges no platform fee on inference. That includes bring-your-own-key requests. You pay Google's rates for Gemini tokens, nothing extra to Vercel for routing them.

This positions AI Gateway as infrastructure rather than a reseller. The monetization presumably comes from Vercel's hosting and deployment products, not from taking a cut of AI spend.

Testing before committing

Vercel offers a model playground where you can try Gemini 3.6 Flash without writing code. The Gateway also publishes a model leaderboard tracking the most popular models by token volume across all Gateway traffic. It's a useful signal for which models the broader developer community actually uses in production, not just which ones get press coverage.

Where this fits in the broader Gemini rollout

Google has been shipping Gemini models at a steady clip, though the flagship 2.5 Pro remains unavailable in many contexts. The Flash tier models, Vercel's new additions, emphasize speed and cost over raw capability. They're the models you reach for when latency matters more than peak intelligence.

For teams already on Vercel's platform, adding these models is low friction. For teams evaluating AI Gateway against alternatives like direct API calls or other routing layers, the no-markup pricing and integrated observability make a reasonable case.

Frequently Asked Questions

What's the difference between Gemini 3.6 Flash and 3.5 Flash-Lite?

Gemini 3.6 Flash targets higher-quality output for coding and web development with improved token efficiency. Flash-Lite is optimized for subagent tasks, handling scoped parts of larger agentic workflows at lower cost.

Does Vercel charge extra for AI Gateway?

No. Vercel states AI Gateway reflects provider pricing with no markup and charges no platform fee on inference, including bring-your-own-key requests.

Can I use my own Google API key with Vercel AI Gateway?

Yes. AI Gateway supports bring-your-own-key requests, letting you use your existing Google Cloud credentials while still getting Gateway features like usage tracking and failover.

How do I switch my existing Vercel AI app to use Gemini 3.6 Flash?

Change the model parameter in your streamText call to 'google/gemini-3.6-flash'. No other code changes required if you're already using the AI SDK.