First things first: we still offer, and always plan to offer, multiple ways to use AI in Zed without paying us.
So what's changing? We're moving Zed AI from prompt-based limits to token-based pricing. For new users, this pricing is live today. For current users, we're providing advance notice and migration will happen over the next three months. We’re also adding GPT-5 (and mini/nano), Gemini 2.5 Pro, and Gemini 2.5 Flash to Zed’s hosted offering, in addition to the Anthropic models already available.
Why? We want our pricing to reflect the real cost of running AI. Token-agnostic prompt structures obscure the cost and are rife with misaligned incentives. This change lets us invest sustainably in the editor features that make Zed fast and reliable. Our short-term revenue growth strategy is to sell enterprise features to businesses, and our long term vision is to earn money by fundamentally improving how developers collaborate. We have to charge for AI features which we get charged for, but it's never been our plan to build a business on LLM token math.
$20$10per month
We've simplified our pricing and reduced costs while adding access to more AI models.
Your Zed Pro bill may go up or down based on how you use AI. That's why we're giving Zed Pro customers 3 months notice before migration to the new pricing structure, and we have invested heavily in alternative options for using AI in Zed:
You can also disable AI in Zed entirely if you prefer to keep your editor free of AI features.
Current pricing is too expensive for Zed. LLM bills have become our biggest expense, and more paying customers translates to more money lost. We could spend more resources trying to find the right "unlimited" price point that would balance out costs, but it's much simpler to switch to the obvious pricing model of charging based on what providers charge us. This lets us keep our focus where it belongs: building the world's best code editor.
Prompt-based pricing is decoupled from user value. In our current system, asking Sonnet to fix a typo costs the same as a complex multi-file refactor. Token-based pricing means you pay for what you actually use, plus a 10% markup that covers Zed’s infrastructure, support costs, and the increased rate limits available for Zed Pro users compared to average BYOK users.
Better incentive alignment, reduced complexity. With token-based pricing, you can add as much or as little context as you want. The only limit is the cost you decide is worth it, not an arbitrary prompt cap. We ourselves would get confused about when “burn mode” was needed, and what usage should or shouldn’t cost a prompt. It also simplifies the addition of new models to Zed Pro: we plan to maintain “list price + 10%” for any new model we add to Zed’s hosted offering.
Subsidize with specificity. Our old free plan subsidized inference without a concrete goal. Changing our pricing helps us focus resources on targeted programs like discounts for students that are better aligned with Zed’s mission. Every user can still use our 14-day trial to evaluate Zed Pro, even if you trialed Zed Pro in our old pricing structure.
If you're an existing customer, you should already have received an email with your specific migration details. Here's the timeline again for reference:
Check out similar blogs from the Zed team.
You can try Zed today on macOS, Windows, or Linux. Download now!
If you're passionate about the topics we cover on our blog, please consider joining our team to help us ship the future of software development.