Use the powerful GPT-4, Llama-3 and more AI models on Raycast, for FREE - no API Key required.
Homepage · Privacy Policy
"If you like the extension, please consider giving it a ✨star✨ tysm!" - the developer, probably
- Install the Raycast app.
- Currently, the extension is only available for macOS, as it requires Raycast. Windows support is in progress.
This extension is currently not available on the Raycast Extension store, but installation from source is extremely simple.
- Download the source code from the latest release, or clone the repository.
- Navigate to the directory, and open a Terminal window at the downloaded folder.
- Run
npm ci --productionto install required dependencies. - (Optional) Run
pip install -r requirements.txtto install Python dependencies. These are required for some features, e.g. web search. - Run
npm run devto build and import the extension.
The extension, and its full set of commands, should then show up in your Raycast app.
Please open an issue if any unexpected problems occur during installation.
There is built-in support for updating within the extension itself! Simply run the "Check for Updates" command in the extension, and it will take care of the update process for you. Furthermore, you can also enable the "Automatically Check for Updates" feature in the preferences.
In the command line, run git pull, npm ci --production and npm run dev (in that order).
You might want to update manually if the automatic update doesn't work (please also open a GitHub issue if this is the case); updating manually also allows you to fetch and view the latest changes to the source code.
▶️ ️ Streaming support - see messages load in real-time, providing a seamless experience.- ⚡ Ask anything from anywhere - with 18 commands available, there's something for you no matter what you need.
- 💪 Support for many providers & models (more info below!)
- 💬 Chat command - interact with the AI in a conversation, and your chat history will be stored in the extension.
- 🌐 Web search - let GPT search the web for the latest information.
- 📄 File upload - you can upload image, video, audio and text files to the AI. (only available for a few providers, more to come!)
- 🎨 Image generation capabilities - imagine anything, and make it reality with state-of-the-art models.
- ✏️ Custom AI Commands - create your own commands with custom prompts!
| Provider | Model | Features | Status | Speed | Rating and remarks by extension author |
|---|---|---|---|---|---|
| Nexra | gpt-4o (default) |  |
Very fast | 8.5/10, the best performing model. | |
| Nexra | gpt-4-32k | |
Medium | 6.5/10, no streaming support but otherwise a great model. | |
| Nexra | chatgpt |  |
Very fast | 7.5/10 | |
| Nexra | Bing | |
Medium | 8/10, GPT-4 based with web search capabilities. | |
| Nexra | llama-3.1 | |
Fast | 7/10 | |
| Nexra | gemini-1.0-pro | |
Fast | 6.5/10 | |
| DeepInfra | meta-llama-3.3-70b | |
Fast | 8.5/10, recent model with large context size. | |
| DeepInfra | meta-llama-3.2-90b-vision | |
Fast | 8/10, recent model with vision capabilities. | |
| DeepInfra | meta-llama-3.2-11b-vision | |
Very fast | 7.5/10 | |
| DeepInfra | meta-llama-3.1-405b |  |
Medium | 8.5/10, state-of-the-art open model, suitable for complex tasks. | |
| DeepInfra | meta-llama-3.1-70b | |
Fast | 8/10 | |
| DeepInfra | meta-llama-3.1-8b | |
Very fast | 7.5/10 | |
| DeepInfra | llama-3.1-nemotron-70b | |
Fast | 8/10 | |
| DeepInfra | WizardLM-2-8x22B | |
Medium | 7/10 | |
| DeepInfra | DeepSeek-V2.5 | |
Fast | 7.5/10 | |
| DeepInfra | Qwen2.5-72B | |
Medium | 7.5/10 | |
| DeepInfra | Qwen2.5-Coder-32B | |
Fast | 7/10 | |
| DeepInfra | QwQ-32B-Preview | |
Fast | 7.5/10 | |
| Blackbox | custom model | |
Fast | 7.5/10, very fast generation with built-in web search ability, but is optimized for coding. | |
| Blackbox | llama-3.1-405b | |
Fast | 8.5/10 | |
| Blackbox | llama-3.1-70b | |
Very fast | 8/10 | |
| Blackbox | gemini-1.5-flash | |
Extremely fast | 7.5/10 | |
| Blackbox | gpt-4o | |
Very fast | 7.5/10 | |
| Blackbox | claude-3.5-sonnet | |
Fast | 8.5/10 | |
| Blackbox | gemini-pro | |
Fast | 8/10 | |
| DuckDuckGo | gpt-4o-mini | |
Extremely fast | 8/10, authentic GPT-4o-mini model with strong privacy. | |
| DuckDuckGo | claude-3-haiku | |
Extremely fast | 7/10 | |
| DuckDuckGo | meta-llama-3.1-70b | |
Very fast | 7.5/10 | |
| DuckDuckGo | mixtral-8x7b | |
Extremely fast | 7.5/10 | |
| BestIM | gpt-4o-mini | |
Extremely fast | 8.5/10 | |
| Rocks | claude-3.5-sonnet | |
Fast | 8.5/10 | |
| Rocks | claude-3-opus | |
Fast | 8/10 | |
| Rocks | gpt-4o | |
Fast | 7.5/10 | |
| Rocks | gpt-4 | |
Fast | 7.5/10 | |
| Rocks | llama-3.1-405b | |
Fast | 7.5/10 | |
| Rocks | llama-3.1-70b | |
Very fast | 7/10 | |
| ChatgptFree | gpt-4o-mini | |
Extremely fast | 8.5/10 | |
| AI4Chat | gpt-4 | |
Very fast | 7.5/10 | |
| DarkAI | gpt-4o | |
Very fast | 8/10 | |
| Mhystical | gpt-4-32k | |
Very fast | 6.5/10 | |
| PizzaGPT | gpt-4o-mini | |
Extremely fast | 7.5/10 | |
| Meta AI | meta-llama-3.1 | |
Medium | 7/10, recent model with internet access. | |
| Replicate | mixtral-8x7b | |
Medium | ?/10 | |
| Replicate | meta-llama-3.1-405b | |
Medium | ?/10 | |
| Replicate | meta-llama-3-70b | |
Medium | ?/10 | |
| Replicate | meta-llama-3-8b | |
Fast | ?/10 | |
| Phind | Phind Instant | |
Extremely fast | 8/10 | |
| Google Gemini | auto (gemini-1.5-pro, gemini-1.5-flash) | |
Very fast | 9/10, very good overall model but requires an API Key. (It's free, see the section below) | |
| Google Gemini (Experimental) | auto (changes frequently) | |
Very fast | - | |
| GPT4Free Local API | - | |
- | allows access to a large variety of providers. read more | |
| Ollama Local API | - | |
- | allows local inference. read more | |
| Custom OpenAI-compatible API | - | |
- | allows you to use any custom OpenAI-compatible API. read more |
📄 - Supports file upload. Note: By default, all providers support basic file upload functionality for text-based files, like .txt, .md, etc.
¹: Supports images only.
- Google Gemini: An API Key is required to use this model. You can get one completely for free:
- Go to https://aistudio.google.com/app/apikey
- Sign in to your Google account if you haven't done so.
- Click on "Create API Key" and follow the instructions there.
- Copy the API Key and paste it into the corresponding box in the extension preferences.
The rate limit for Google Gemini is 1500 requests per day (as of the time of writing). This should be much more than enough for any normal usage. If your use case needs an increased rate limit, you can even create multiple API Keys with different Google accounts; separate them with commas in the preferences.
- Google Gemini: This provider supports File upload functionality, as well as the Ask About Screen Content command! To upload a file in AI Chat, press Command-Enter or select "Compose Message" from the actions. Then, simply click on the upload button to get started.
Let GPT decide to search the web for information if it does not have enough knowledge or context. Uses DuckDuckGo search, fast and free.
Enabling web search is fast and easy. Go to the extension preferences, and the "Web Search" option will be available. There are 4 options:
- Disabled (default)
- Automatic: Enable Web Search only in AI Chat. GPT will automatically decide when to use it.
- Balanced: Use Web Search in every query for AI commands¹, and automatically in AI Chat. This is basically an extension of the "Automatic" option.
- Always: Always use Web Search for every query, both in AI Chat and in commands¹.
¹: Commands that support Web Search are: Ask AI, Ask About Selected Text, Explain. Other commands will not use Web Search.
Web Search is also available in the following commands:
- Custom AI Commands: You can enable Web Search for each command individually.
- AI Chat: You can enable Web Search for each chat individually.
- AI Presets: You can enable Web Search for each preset individually.
Let GPT automatically come up with a name for the current chat session after you send the first message. For example, this is similar to what the ChatGPT web UI does.
Let the extension automatically check for updates every day. If a new version is available, you will be notified, along with the option to update the extension with a single click.
Enable more persistent storage of the extension's data, like AI Chat data or Custom Commands. This will back up a copy of this data to files on your computer. Useful for saving large amounts of data. Note: With this option off, your data is already well preserved. Do not enable this if you have sensitive data.
Show a cursor icon when the response is loading - cosmetic option only.
Allows GPT to execute Python code locally. The model has been instructed to strictly only produce safe code, but use at your own risk!
Only models with function calling capabilities support this feature. Currently, this includes only selected DeepInfra models.
- I submitted it when the extension was in its early stages, but since Raycast is quite conservative about adding AI extensions, it was rejected because of concerns over the use of third-party APIs. I think this is a valid concern, and I'd like to be very transparent about where your data is going to, so I'll be writing all the details in the project homepage really soon. But do rest assured that I use safe APIs from legitimate websites; I also update the extension very frequently so all the providers I use will be up-to-date.
- Thus, the extension will have to be installed from source. Regarding this, I apologize as it's indeed more complicated than downloading it from the store. I have tried my best to make the installation process quick and streamlined - please do provide feedback on whether it was simple enough!
- raycast-g4f is completely free! I strongly believe that such AI features, especially in an awesome productivity tool like Raycast, shouldn't be locked behind a paywall.
- Raycast AI definitely has more model options, but the quality of AI in this extension is, in my opinion, comparable. Some models available here are gpt-4, gpt-4o, Claude 3.5 Sonnet and Google Gemini, as well as capable open-source models like Llama 3.1. If you still doubt it, I'd encourage you to give the extension a try! :)
- Because the developer API that Raycast provides is limited, it's not possible to replicate the Raycast AI interface exactly. The components I can use are very simple, and a lot of the features in built-in Raycast commands are not available to extension developers.
- However, I've spent a lot of effort trying to make the UI really intuitive, and I'm always open to feedback on how to improve it!
- Sometimes third party providers can be slow or unresponsive. If you're experiencing this issue, please try again in a few minutes; or if the problem persists, please try switching to another provider.
- If you've tried various providers and the issue still persists, please open an issue on GitHub!
- I welcome all contributions! If you have an idea for a new feature, or if you've found a bug, please open an issue.
- If you'd like to contribute code, please open a pull request, and I'll make sure to review it as soon as possible.
License: GPLv3. Full license is found in LICENSE.txt.
The code base is derived from Raycast Gemini by Evan Zhou.
Third-party libraries used for generation:
(Both packages are maintained by the extension author.)
Some of the code in this repository was inspired or ported from the original gpt4free project (written in Python).