Raycast Pro AI Models: Full List, Plans, and Use Cases
Published April 22, 2026 • 9 min read
If you're wondering what AI models Raycast Pro supports, the short answer is: a lot. Raycast doesn't give you one AI assistant, it gives you access to over a dozen AI providers inside a single Mac launcher, all triggered from the same keyboard shortcut. This guide covers which models are actually available, what each tier costs to unlock, and which model is worth using for your specific workflow on a given day.
The model list has grown significantly in 2026. Raycast now hosts everything from OpenAI's GPT-5 family to Anthropic Claude, Google Gemini, Mistral, DeepSeek, Perplexity Sonar, xAI Grok, Meta Llama, and Raycast's own Ray-1 model. That's a lot to sort through without a guide, so what follows is a model-by-model breakdown with plan requirements, configuration steps, and use-case recommendations.
If you're still evaluating whether to subscribe, a current discount offer through Raycast Pro Discount 2026 is worth checking before you commit to full price, details at the end.
What AI Models Does Raycast Pro Support, Full List by Provider
The breadth of Raycast's AI integrations is one of its strongest selling points. Rather than being locked into a single provider, you get a multi-model picker that spans every major LLM family. For a concise overview of the product's AI capabilities, see Raycast's AI features page.
OpenAI's GPT Family in Raycast
OpenAI has the deepest integration inside Raycast. The full lineup includes GPT-5, GPT-5.1, GPT-5 Reasoning, GPT-5 mini, GPT-5 nano, GPT-4.1, GPT-4.1 mini, GPT-4o, GPT-4 Turbo, and GPT-4. That range gives you both cutting-edge reasoning models and lighter, faster variants in the same interface, no need to switch to a heavy model for a quick lookup. Our Raycast and ChatGPT walkthrough goes deeper on the OpenAI-specific workflow.
Anthropic Claude, Google Gemini, Perplexity, and xAI
Anthropic's Claude 3.7 Sonnet is the flagship Claude option in Raycast. In practice, it outperforms other available models on extended coding sessions and structured writing tasks, largely because of its 200,000-token context window. If Claude is your main driver, our Raycast + Claude guide covers the integration in detail. Google Gemini is also a supported provider. Perplexity's three Sonar models, Sonar, Sonar Pro, and Sonar Reasoning Pro, stand out because they pull live web context into their responses, making them genuinely useful for research. xAI's Grok-4.1 Fast and Grok-4.1 Fast Reasoning round out this group as newer additions with strong reasoning speeds.
Mistral, Meta Llama, DeepSeek, and Raycast's Own Models
The full list extends further. The Mistral family covers Mistral Large, Mistral Medium, Mistral Small 3, Mistral Nemo, and Codestral. Meta's Llama 4 Scout and Llama 3.3 70B are available via Groq. DeepSeek-V3 and DeepSeek-R1 come through Together AI. Qwen3 variants and Moonshot AI's Kimi K2.5 fill out the third-party options.
Worth calling out separately: Raycast's own Ray-1 and Ray-1 mini models are native to the platform, meaning they're optimized for the launcher's workflow patterns rather than being a generic hosted API. If you came here looking for a specific model name, it's almost certainly on this list, or accessible via the custom providers covered later.
Which Plan Unlocks Which AI Models in Raycast Pro
The plan structure is straightforward once you see the three tiers laid out clearly. Most confusion comes from not knowing which models sit behind the Advanced AI add-on versus what's included in base Pro. For a full billing breakdown, our Raycast Pro pricing guide lays out every tier side-by-side.
What the Free Tier Gives You (and Its Limits)
The free plan includes 50 AI trial messages across basic models. That's enough to test the interface and get a feel for how AI integrates into Raycast's launcher workflow, but it won't hold up under daily use. No advanced models, no unlimited access, no Cloud Sync. Think of it as a preview, not a working setup.
What Raycast Pro Covers at $8/Month
The Pro plan runs $8 per user per month on annual billing, or $10 on monthly billing. It unlocks unlimited messages on Raycast's basic AI models, Cloud Sync, unlimited Clipboard History, and the full productivity feature set including custom themes and notes. Rate limits apply, 50 requests per minute, 300 per hour, but those are rarely hit in normal use. The plan also includes a 14-day free trial, which removes the risk from signing up before you're fully committed. For the official plan details, see the Raycast Pro page.
When the Advanced AI Add-On Matters
The top-tier models sit behind the Advanced AI add-on, which costs an additional $8 per month on top of Pro. This tier unlocks GPT-4, GPT-4 Turbo, Claude 3.7 Sonnet, and equivalent premium models from other providers. To put the cost in context: accessing GPT-4 through Raycast's Advanced AI add-on still costs less than a standalone ChatGPT Plus subscription at $20 per month, and you get every other model on the list simultaneously.
If you want to test both tiers before paying full price, the promotional pricing available through Raycast Pro Discount 2026 applies an 80% discount automatically at checkout, no coupon code required, covering both the base Pro plan and the path to the Advanced AI add-on.
How to Switch Models and Configure Your AI Setup
Raycast's AI configuration is centralized in one place, which makes it easy to manage once you know where to look.
Setting Your Default Model for Quick AI and AI Chat
Open Raycast Settings (Cmd+,) and navigate to the AI tab. From here, you set separate default models for Quick AI commands and AI Chat conversations, these can be different models, which is useful if you want a fast lightweight model for quick lookups and a more capable one for extended conversations. Switching mid-conversation is also possible: use the model dropdown at the top of the chat window, and press Shift+Cmd+, to save that preference for the current conversation. For step-by-step screenshots and additional settings, consult Raycast's AI documentation. Our Raycast AI commands guide covers how to package model choices into reusable shortcuts.
Adding Your Own API Keys for More Control
The Custom API Keys section in Settings → AI lets you bring your own keys from OpenAI, Anthropic, or other supported providers. When you add a key, Raycast uses it instead of its own quota, useful if you have existing API credits or want to operate outside Raycast's rate limits. A key icon appears next to models unlocked this way, so you always know which requests go through your account versus Raycast's shared allocation.
Using Auto Model to Let Raycast Choose
The Auto Model feature, currently experimental, automatically selects the most appropriate model based on the type of task you're running. For users who don't want to think about model selection, this is a practical default. You can also run the "Manage Models" command to hide providers you never use, which keeps the model picker clean and fast to navigate. If you use the same prompts every day, pair this with Raycast AI Presets to lock in model + system prompt combinations per workflow.
Picking the Right Raycast AI Model for Your Mac Workflow
With this many models available, the practical question becomes: which one should you actually use? The answer depends on what you're doing. For a structured head-to-head of GPT-4o vs Claude vs Gemini specifically inside Raycast, see our Raycast AI models comparison.
Coding and Developer Tasks
GPT-5.1 and its Codex variants are the frontrunners for code generation, debugging, and agentic coding tasks. GPT-5 and GPT-5.1 handle complex multi-step reasoning well, exactly what you need when tracing bugs across multiple files. Claude 3.7 Sonnet is the other strong contender here; its 200,000-token context window makes it especially useful when working with large codebases. Both models pair well with Raycast's native developer extensions like GitHub and terminal integrations.
Multilingual Work and Writing
Mistral Large is the standout for multilingual tasks, with strong reasoning across non-English languages that most other models don't match at the same speed tier. Claude 3.7 Sonnet performs consistently well for long-form writing and summarization, producing outputs that require less editing before they're usable. For research-heavy writing where you need current information, Perplexity's Sonar Pro pulls real-time web context into the response, something the other models on this list can't do natively.
Fast, Lightweight Responses for Everyday Use
When you're triggering AI dozens of times a day for quick commands, reminders, and single-question lookups, inference speed matters more than raw capability. GPT-4o mini, Groq-hosted Llama 3.1 8B, Mistral Nemo, and xAI Grok-3 Mini are the practical choices here. These models prioritize low latency, and the speed difference is noticeable when you're moving between the launcher and your actual work throughout the day.
Running Local and Custom Models in Raycast
For users with privacy requirements or existing API credits at other providers, Raycast's custom model support is one of its most underappreciated features.
Setting Up Ollama for Private, Offline AI
Download Ollama and move it to your Applications folder. Once it's running, you can add local models directly from Raycast Settings → AI → "Add Ollama Model." Over 100 models are supported, including the full Llama family ranging from small 135M-parameter variants to 671B-parameter versions. This option works on the free plan, costs nothing beyond your hardware, and keeps all data local, making it the right configuration for anyone with strict data privacy requirements.
Connecting Replicate and Other External Providers via YAML
Raycast's Custom Providers feature, Replicate and custom model support in Raycast included, supports any OpenAI-compatible API endpoint through a YAML config file at ~/.config/raycast/ai/providers.yaml. The structure defines a provider ID, base URL, API key, and model list. A Replicate entry, for example, would use Replicate's API base URL and a model ID like meta/llama-3.1-8b-instruct. Refer to Replicate's official models documentation for model IDs and endpoint details. If you prefer not to edit YAML directly, the Raycast AI Custom Providers extension offers a GUI-based alternative for managing providers without touching the config file.
When to Go Custom Versus Sticking with Built-In Models
Built-in models cover the vast majority of use cases. Custom providers and local setups make sense in a few specific situations:
- You have privacy requirements that prevent sending data to third-party APIs
- You have existing API credits at a provider and want to route them through Raycast
- You need a specific model that Raycast doesn't include by default
Outside those scenarios, the built-in model list is comprehensive enough that custom setup adds overhead without meaningful benefit.
Testing Every Model Without Overpaying
To recap what AI models Raycast Pro supports: you're looking at GPT-5, Claude, Gemini, Mistral, Llama, DeepSeek, Perplexity Sonar, xAI Grok, and a dozen more, all accessible from a single keyboard shortcut without managing a separate app for each. The plan structure is clear: base Pro covers most users with unlimited basic AI access, and the Advanced AI add-on is worth it if you rely on top-tier reasoning models for coding or complex analysis daily. Not sure which tier fits? Start with our discount landing page for the live offer.
The 14-day free trial removes the financial risk entirely. You get full access to evaluate the workflow before a single dollar leaves your account, enough time to run real tasks and see whether Claude 3.7 Sonnet or GPT-5 actually improves your output compared to the base models. Raycast Pro Discount 2026 then applies an 80% discount automatically at checkout after the trial, with no coupon code required. It's the lowest-friction way to get started with every model covered in this article, including the advanced tier, at a fraction of standard pricing.
Frequently Asked Questions
Which AI models does Raycast Pro support?
Raycast Pro supports over a dozen AI providers, including OpenAI's GPT-5, GPT-5.1, GPT-5 Reasoning, GPT-4.1, GPT-4o, Anthropic Claude 3.7 Sonnet, Google Gemini, Perplexity Sonar, xAI Grok-4.1, Mistral (Large, Medium, Small, Nemo, Codestral), Meta Llama 4 Scout, DeepSeek, Qwen3, Moonshot Kimi K2.5, and Raycast's native Ray-1 models — all from a single keyboard shortcut.
What's the Advanced AI add-on?
The Advanced AI add-on costs an additional $8 per month on top of Raycast Pro and unlocks top-tier models like GPT-4, GPT-4 Turbo, Claude 3.7 Sonnet, and equivalent premium models from other providers. Together with Pro, it still comes in below a standalone ChatGPT Plus subscription while giving you every supported model in one interface.
Can I use my own API keys?
Yes. Settings → AI → Custom API Keys lets you bring your own keys from OpenAI, Anthropic, or other supported providers. When a key is configured, Raycast routes requests through your account instead of its own quota, which is useful if you have existing credits or want to bypass Raycast's shared rate limits. A key icon appears next to models unlocked this way.
Does Raycast AI work offline?
The built-in hosted models require an internet connection, but Raycast supports local, offline AI through Ollama. Install Ollama, then add local models via Settings → AI → Add Ollama Model. Over 100 models are supported, including the full Llama family, and this option is available on the free plan with all inference running locally.