Raycast AI statistics start with the monthly allowance. Raycast Pro includes 500 AI credits, Pro+ 3,000 and Max 7,500, and Raycast Free doesn’t include AI. Raycast moved its plans from rate-limited to usage-based and added Max at $50 per month, so the old request limits no longer describe what a Raycast v2 subscriber gets.
Credits measure AI usage, not a fixed number of messages, and depend on the model and how much text it reads and writes. Every figure below comes from Raycast’s own pricing page, billing and AI manuals, models page, changelog, blog, FAQ and developer docs, read in October 2026. They cover plan limits, credits per dollar, the request caps that still apply on Raycast v1, credit expiry, Teams pricing, the model lineup and the ways to bring your own model.
Key Takeaways
- On monthly billing, Pro+ and Max give 150 AI credits per dollar, 3 times the Pro rate.
- Raycast’s models page lists 48 models from 12 providers, led by OpenAI with 13.
- Before December 17, 2026, Raycast v1 users who run out of credits keep 50 requests per minute and 300 per hour on Pro models.
- Each monthly credit grant expires 60 days after it’s granted, and free trial credits expire after 7 days.
- Top-up packages run from 500 to 10,000 credits and expire 12 months after purchase.
- Subscribers already on a yearly individual Pro+ plan on September 11, 2026 get 4,500 credits a month instead of 3,000 for the rest of their current annual term.
- Bring Your Own Key accepts API keys from 4 providers: Anthropic, Google, OpenAI and OpenRouter.
Editor’s Choice
- Max plan allowance: 7,500 AI credits a month for $50.
- Pro price on annual billing: $8 a month, or $96 billed annually.
- Teams Max price: $65 per member a month.
- Largest top-up package: 10,000 credits.
- Top-tier models for Pro+ and Max: 10 models.
- Automations per device: up to 20.
Raycast AI Statistics: Limits and Credits by Plan
- AI features aren’t included in Raycast Free, and one subscription covers all three platforms: macOS, Windows and iOS.
- Raycast Pro costs $10 a month or $96 a year and includes 500 monthly AI credits, listed at $8 a month billed annually.
- Raycast Pro+ costs $20 a month or $192 a year with 3,000 credits.
- Raycast Max costs $50 a month or $480 a year with 7,500 credits.
- Top-tier models are exclusive to Raycast Pro+ and Raycast Max plans, and the comparison table on Raycast’s pricing page shows the middle tier as Plus with 3,000 credits.
- Pro now includes models like Claude Sonnet 5 or GPT-5.6 Terra that used to require Advanced AI.
- Long conversations, attachments, tool calls, and regenerated or edited messages all add up.
Plan | Monthly price | Yearly price | Monthly AI credits | AI models |
|---|---|---|---|---|
Raycast Free | $0 | $0 | Not included | Not included |
Raycast Pro | $10 | $96 | 500 | Fast and powerful models |
Raycast Pro+ | $20 | $192 | 3,000 | All models, including top-tier |
Raycast Max | $50 | $480 | 7,500 | All models, including top-tier |
Source: Raycast Billing, Raycast Pricing, October 2026
Credits apply only to inference provided by Raycast AI, and models you bring yourself follow your own provider’s limits instead.
About This Data
- Compiled from 16 sources: 9 Raycast manual and developer pages, 5 Raycast website pages and 2 Raycast blog posts.
- Sources run from May 2023 to October 2026.
- Only figures Raycast publishes about its own plans, models and limits qualified.
- Figures are reviewed on a rolling basis and updated when Raycast changes its plans.
Raycast AI Credits per Dollar
- On monthly billing, Pro buys 50 credits per dollar, while Pro+ and Max both buy 150.
- Yearly Pro+ works out to $16 a month, and yearly Max to $40 a month.
- Paying yearly saves $24 a year on Pro, $48 on Pro+ and $120 on Max.
View the data
Key finding: Raycast’s billing prices give Pro+ 3 times as many AI credits per dollar as Pro on monthly billing. Pro gives the fewest credits per dollar of the three plans, so a Pro subscriber who keeps running out gets more AI per dollar by moving up a tier, and Max then adds volume at the Pro+ rate.
Yearly billing lifts the rate to 62.5 credits per dollar on Pro and 187.5 on Pro+. Max matches Pro+ at 187.5.
Recent Developments
- October 16, 2026: Raycast 2.8 added browsing of installed extension skills in AI settings and clarified when you’ve reached the weekly Claude Fable usage limit.
- October 9, 2026: Raycast 2.7 made AI requests smaller, with initial context 15% smaller for normal queries.
- October 5, 2026: Raycast 2.6 reworked the Auto model with Lightweight, Balanced and Flagship preferences, and Flagship requires a Pro+ or Max subscription.
- September 10, 2026: Raycast moved its plans from rate-limited to usage-based, while all subscription prices stayed the same.
- September 10, 2026: Existing subscribers received two full monthly credit packs at once, and these credits expire after 30 days.
- September 10, 2026: Users can connect a Claude or ChatGPT account to any paid Raycast plan and use those models in AI Chat, Quick AI and AI Commands.
Raycast Advanced AI Limits on Raycast v1
- Credits apply to Raycast v1 users on Raycast Pro or Raycast Pro+ too, with a transition period to give them time to update.
- Before December 17, 2026, Raycast tracks usage and credit balance, but nothing changes in practice.
- At 0 credits, Pro models stay at 50 requests per minute and 300 requests per hour, shared across all Pro models.
- From December 17, 2026, Raycast AI on v1 stops when credits reach 0 and asks you to update Raycast or wait for your next monthly credits.
| Model group on Raycast v1 | Request limit at 0 credits |
|---|---|
| Pro models | 50 per minute, 300 per hour |
| o3-mini and o4-mini | 150 per 24 hours |
| Advanced AI models (Pro+) | 75 per 3 hours, 150 per 24 hours |
| Top-tier models | 50 per week per model |
Source: Raycast AI Credits manual, October 2026
New models and top-tier models added to Raycast Pro+ and Raycast Max from September 10, 2026, are only available in Raycast v2.
Raycast AI Credit Expiry and Rollover
- Unused monthly credits roll over for one extra billing cycle, and each monthly grant expires 60 days after it’s granted.
- Free trial credits expire 7 days after they’re granted.
- Top-ups come in packages of 500, 2,500, 5,000 and 10,000 credits and expire 12 months after purchase, and the largest package equals 20 times the Pro monthly allowance.
- Credits are used in order of expiry date, which usually means monthly credits are used before top-up credits.
- Top-ups aren’t available during the Raycast Pro free trial, and on Teams plans they are per member and aren’t pooled.
- Cancelling doesn’t change when credits expire, but an active paid subscription is required to use them.
View the data
When both monthly and top-up credits are exhausted, Raycast AI stops until the next allowance arrives or you purchase a top-up. A provider rate limit is different, and buying credits does not remove it.
Raycast Teams AI Pricing per Member
- Teams Pro includes 500 monthly AI credits per member, Teams Pro+ 3,000 and Teams Max 7,500.
- Billed yearly, Teams Pro costs $144 per member, Teams Pro+ $240 and Teams Max $624.
- Teams Max costs $65 per member a month, $15 more per member each month than individual Max.
- Credits are per member and aren’t pooled across the team, and there’s no minimum or maximum number of seats.
- The AI Control Center, with an organization-wide toggle, bring-your-own keys, an AI provider allow-list and a custom AI provider, is included only on the custom-priced plan in the pricing page’s comparison table, not on Free, Pro, Plus, Max or the Teams plans.
View the data
On the pricing page, Teams Pro is listed at $12 per user a month billed annually ($144 a year).
Raycast AI Bonus Credits for Yearly Subscribers
- Subscribers already on a yearly individual Pro or Pro+ plan on September 11, 2026 get 50% more credits each month for the rest of their annual term.
- Pro+ goes from 3,000 to 4,500 credits a month.
- That adds 1,500 credits a month on Pro+ and 250 on Pro.
- New subscribers and subscriptions on monthly billing keep the standard credits.
View the data
The higher amount applies from the next credit reset, because everyone already received double credits as the launch bonus.
Raycast AI Models by Provider
- OpenAI supplies 13 of the models on Raycast’s AI models page, and Anthropic 8.
- 42 of the 48 models list a 200,000-token context window.
- GPT-OSS 20b and GPT-OSS 120b (Groq) and the Gemma 4 model (Vercel) list 131,000 tokens, Codestral 128,000, and Sonar and Sonar Reasoning Pro 127,000.
- 14 of the models are served through Vercel, including Kimi K3, DeepSeek V4 Pro and Qwen3.8 Max.
| Provider | Models listed |
|---|---|
| OpenAI | 13 |
| Anthropic | 8 |
| 5 | |
| Mistral | 4 |
| Perplexity | 3 |
| Z.ai | 3 |
| DeepSeek | 3 |
| xAI | 2 |
| Meta | 2 |
| Moonshot AI | 2 |
| Thinking Machines | 2 |
| Alibaba | 1 |
Source: Raycast AI Models page, October 2026
Raycast AI Top-Tier Models
Top-tier models are exclusive to Raycast Pro+ and Raycast Max plans, and the list will evolve as new frontier models are released. The current list holds 10 models, 5 from OpenAI and 5 from Anthropic.
| Provider | Top-tier models |
|---|---|
| OpenAI | GPT-6.1 Sol, 6 Astra, 6 Sol, 5.5, 5.5 Instant |
| Anthropic | Claude Opus 5.5, Fable 5.1, Opus 5, Opus 4.8, Opus 4.7 |
Source: Raycast AI Credits manual, October 2026
Raycast Bring Your Own Key, Subscription and Local Models
- Bring Your Own Key (BYOK) covers 4 API providers, and Bring Your Own Subscription covers 2 subscriptions, Claude and ChatGPT.
- Bring Your Own Key (BYOK) lets you connect your own API key from Anthropic, Google, OpenAI, or OpenRouter, with one API key per provider.
- For Anthropic, Google and OpenAI, only models available in Raycast AI work through BYOK, while an OpenRouter key unlocks OpenRouter’s full model catalog.
- Bring Your Own Subscription uses a Claude or ChatGPT subscription inside Raycast, and usage counts towards that subscription instead of Raycast AI credits.
- Claude and ChatGPT both have rolling limits, typically a five-hour window and a weekly cap, set by the provider.
- Accounts that only have API access don’t qualify for Bring Your Own Subscription.
- Local models run through Ollama, work without an internet connection and don’t use AI credits for the conversation itself.
- Your own API key, a custom provider, or a local model now requires a paid Pro plan.
Option | Providers | Who bills the model | Conversation billed to Raycast AI credits |
|---|---|---|---|
Bring Your Own Subscription | Claude, ChatGPT | Your Claude or ChatGPT subscription | No |
Bring Your Own Key | Anthropic, Google, OpenAI, OpenRouter | Your provider API account | No |
Custom providers | Your configured provider | The provider, at standard rates | No |
Local models | Ollama | Runs on your own hardware | No |
Source: Raycast AI manual, Raycast AI, October 2026
Raycast says API keys and Claude or ChatGPT subscriptions work with Raycast AI without consuming AI Credits. For anyone who searches for Raycast and OpenRouter, BYOK is the route, and an OpenRouter key through BYOK unlocks OpenRouter’s full model catalog.
Raycast AI Auto Model and Context Savings
- Raycast’s reworked Auto model has three preferences: Lightweight, Balanced, and Flagship.
- Balanced is the default and uses about 70% fewer credits than Flagship.
- Lightweight responds in about 70% less time than Flagship and works well for quick questions.
- Initial context is now 15% smaller for normal queries, and 40% smaller with automatic extension and skill discovery turned off.
- Raycast 2.8 reduced the data sent at the start of a chat by loading extension tools more selectively.
| Auto model preference | What Raycast reports |
|---|---|
| Lightweight | Responds in about 70% less time than Flagship |
| Balanced (default) | Uses about 70% fewer credits than Flagship |
| Flagship | Most capable models, requires Pro+ or Max |
Source: Raycast Changelog, October 2026
Raycast AI Privacy and Data Retention
- Raycast says all models provided by Raycast AI are Zero Data Retention (ZDR), with three exceptions: Mistral AI and xAI retain data for 30 days, and Anthropic retains Claude Fable 5.1 data for 30 days.
- Raycast’s agreements with AI providers prohibit them from using any AI interactions to train their models.
- With a connected subscription, prompts go straight from the provider’s tool on your machine to Anthropic or OpenAI, never through Raycast’s servers.
- Custom provider requests are routed directly from your device to the provider’s servers, without passing through Raycast.
| Provider or model | Data retention |
|---|---|
| Raycast AI models (default) | Zero Data Retention |
| Mistral AI | Rolling 30-day window |
| xAI | 30 days |
| Anthropic (Claude Fable 5.1) | 30 days |
Source: Raycast AI Privacy & Security, October 2026
Why it matters: Anthropic retains data for 30 days on Claude Fable 5.1 and doesn’t offer Zero Data Retention for this model, while other Anthropic models remain Zero Data Retention. Anyone pasting client data into Raycast AI should check which model a chat uses, because one of the top-tier models keeps prompts for 30 days while most of the lineup keeps nothing.
Raycast AI Automations and Extension API
- Automations run an AI prompt on a schedule, and each run is a normal AI Chat message that can use AI Extensions and Skills.
- Automations live on the device where you created them, don’t sync through Cloud Sync, and each device holds up to 20 automations.
- AI Commands turn go-to prompts into keyboard shortcuts to fix grammar, shorten an email or adjust your tone.
- Raycast’s AI API gives extension developers access to AI without requiring API keys, configuration, or extra dependencies.
- If a user doesn’t have Raycast Pro, they are asked whether they want access when an extension calls the AI API.
- The API’s default model is
OpenAI_GPT-5.6_Luna.
| Feature | Limit or requirement |
|---|---|
| Automations per device | Up to 20 |
| Automation sync | Not synced through Cloud Sync |
| Extension AI API access | Raycast Pro |
| Extension AI API default model | OpenAI_GPT-5.6_Luna |
Source: Raycast Automations manual, Raycast API, October 2026
Raycast AI Pricing Timeline
- In 2023, Raycast introduced Raycast Pro and Raycast AI for $8 per month, billed annually, when Raycast’s daily users passed 100,000.
- Before that launch, thousands of people tested the Raycast AI beta, generating over 700 million words.
- Raycast says AI started in 2023 as one question in and one answer out, which made usage easy to estimate.
- AI Chat now takes a task, uses your extensions, writes and runs code and keeps going until it’s finished.
- The cost of supporting these workflows has increased, while Raycast’s subscription prices haven’t.
- The Advanced AI add-on no longer exists and has been replaced by Raycast Pro+, a standalone plan.
View the data
How expensive is a Raycast?
Raycast is free to use, with paid plans that unlock more powerful features and AI across macOS, Windows, and iOS. AI starts with Raycast Pro at $8 a month billed annually, or $10 on monthly billing, and Raycast Max costs $50 a month.
Raycast offers a Pro plan 50% discount to verified students.
What exactly is Raycast?
On top of searching local apps and files, Raycast can be your clipboard manager, text expander, window manager and command palette for all apps. Raycast says it provides a much richer feature set than Apple’s Spotlight and the Alfred launcher.
You can install extensions for Jira, GitHub, and Google Workspace from the Store and control them without opening a browser.
Conclusion
Raycast AI now runs on monthly credits: 500 on Pro, 3,000 on Pro+ and 7,500 on Max, and Max costs $50 a month. Pro+ and Max both work out to 150 credits per dollar on monthly billing, against 50 on Pro. For heavy users, the bigger lever sits outside the credit table: a connected subscription, an API key or a local model takes the conversation off Raycast’s meter.
From December 17, 2026, Raycast AI on v1 stops when credits reach 0. Anyone still on v1 should plan the move to v2 before then, since new models and top-tier models added from September 10, 2026 are only available there.


Leave a comment
Have something to say about this article? Add your comment and start the discussion.