The Gift of AI – 2024 Holiday Edition
Top Tools You Can Use
Introduction
As I finished wrapping presents for Christmas, I realized that the best presents we can share this season are the amazing AI models and AI tools released this year. This article shares as a holiday treat some of the best AI tools you can use, and where and how to get them.
No wrapping is needed for these AI models and AI tools. Rather, they are typically accessible in one of three ways:
Open-source and open-weights AI models are the most freely accessible, both free as in zero-price and free as in freedom to use. They are available to use directly via HuggingFace or download to local use with or Ollama. Open-source AI tools likewise are available to use.
Free tier. Trying tools out for free is more than possible, especially for an occasional user, thanks to free tools or generous free tier options for many AI models and tools.
Subscription-based AI tools. Proprietary AI models and tools monetize via subscriptions, and so require a subscription for full access to top AI model and tool features. Most subscriptions are reasonable priced if you are a professional user.
AI got better across the board this year, so it’s not a surprise that practically everything we mention is either new or vastly upgraded in 2024.
AI Frontier Model Apps
This article won’t cover any AI devices. The Rabbit R1 and Humane Pin didn’t live up to billing and flopped. While we see a lot of promise in Meta’s Ray-Ban AI glasses and Google’s Android XR headsets with Project Astra-like capabilities, let them work out the product and features first. In the meantime, the best in-person access to frontier AI is via smartphone apps.
The three leading AI labs – Google, OpenAI, and Anthropic – all have smartphones apps (both on Apple and Android), desktop apps, and in browser.
The ChatGPT smartphone app now has web search and voice-interaction mode, which combined is extremely useful; you can verbally ask for up-to-date information on news, sports scores or weather, and get a voice report. Their browser app canvas is an excellent feature for writing and coding iteration.
Google’s Gemini smartphone app also can give up-to-date information in voice-interaction mode, such as latest NFL scores, or verbally answering in-depth questions on topics. Hands free AI is very compelling. Now it has Gemini 2.0 Flash Experimental, their best model that is fast to boot. You can use Gemini 2.0’s multimodality to understand and verbally explain a picture or image, such as identifying a household item (it could confirm a specific light bulb type for me) or explaining a chart.
Anthropic’s Claude smartphone app lacks full voice interactivity but takes audio input to generate text output; I don’t find it compelling. The Claude Windows app is more useful. You get the full Claude interface in a Windows app. Claude 3.5 Sonnet has strength in generating proficient writing and code, so it’s a great sidebar tool for AI coding and writing projects.
A free use honorable mention goes to Grok on X, which is now free even to non-paid X users. It also has Flux image generation for free and with little censorship, so make your own memes. A good use-case for Grok is getting Perplexity-like news updates or reports on topics that weblinks and X post sources and references.
Microsoft Copilot is also free and fairly useful way to access GPT-4o. It’s useful for generating images via Dall-E 3 that can be fed into Microsoft designer for touch-up. Back in 2023, Copilot was my number one go-to, but as other options have gotten better, I’ve used it less; we’ll need to see if and when they bring in o1 model. For those not signed up to ChatGPT subscription, it’s a suitable alternative.
Last but not least, how to access the open-source AI models? In my June 2024 article “The Best Ways to Run AI Locally,” I mentioned Ollama, Jan and OpenWebUI. I use Ollama regularly, and use it to access the latest and greatest open source models like Qwen 2.5, Phi-4 and Llama 3.3.
As an aside, Phi-4 14B is a great model that punches way above its weight class; it beats out GPT 4o-mini and Llama 3.3 70B on many benchmarks. It’s worth trying out.
For my own use, I pay $20/month for ChatGPT plus to get their o1 model for reasoning. I’ve found the Google Gemini Advanced worthwhile as well for $20/month; getting access to Gemini 2.0 advanced models makes it even more compelling. Beyond that, I’ll use only free tier and open-source models. I’ve used the free tier of Grok and Claude 3.5 Sonnet as needed, and I have Ollama available for local use of AI models.
Writing and Text Generation
Writing support is now embedded in Google tools (Gmail, Google doc) and Microsoft tools (CoPilot), most of which are available for free. Further, the Claude Artifacts and ChatGPT canvas feature make them useful alternative helpers. Claude now has a writing style editor to tweak its output to be more informal, formal, explanatory, or concise, depending on user preference.
As AI gets better, this leaves pioneers in AI writing support like Grammarly in an uncertain place. Many of their once-unique features are getting subsumed by a plethora of options from ever more powerful AI models. Yet, it still has value as a focused tool. Grammarly has a free tier that is powerful enough for everyday use, helping you draft error-free emails, essays, and social media posts.
For some professional writers, it may be worth the benefits of integration and ease-of-use to get a specific AI tool. For the rest of us, the free tier tools and features should suffice. For example, use embedded Word or Google doc features or adopt a flow to get advice and input from ChatGPT or other AI models.
Code Generation
Our September article “AI for Coding – AI Coding Assistants” highlighted some of the leading AI coding tools: GitHub Copilot, Cursor, Codeium, Aider and Continue. The rapid progress in the world of code generation has continued, with a few highlights worth noting:
Codeium released Windsurf Editor in November, which introduced “AI Flows” that combine the collaboration of Copilots with automation of Agents.
Claude.dev became Cline. It’s a coding assistant implemented as a VS Code extension that supports a variety of AI models. Their Cline 3.0 now uses Model Context Protocol (MCP) to connect to tools and extend its capabilities.
Cursor has been the leading tool in this space, with many developers loving it. However, they don’t have a free tier and there are a number of interesting alternatives, including free tools like Cline and Aider that are worth considering, if you are a developer.
Image, Video and Music Generation
There are now many options for AI image generation, and you can access many of them free:
The original AI image generation model Dall-E 3 is available on ChatGPT and Microsoft Copilot.
Midjourney is perhaps still the leading AI image generation tool, but it lacks a free tier.
Ideogram is great for its rendering of text, remix options, exploration, and has a free tier.
BlackForestLabs released Flux this summer, and it became a hit for its state-of-the-art realism and rendering. You can access Flux image generation via Fal.ai playground and via Grok.
Google’s just-released Imagen 3 is now the highest-resolution AI image generation model, and it’s available via Gemini.
With the recent releases of Sora, Veo 2, Kling 1.6, and Pika 2.0, the quality and options in AI video generation just got much better. With such rapid change, it’s a good time to experiment with what’s available and see what works for you. It will keep improving.
The same is true for music and audio generation, where tools like Suno and ElevenLabs continue to evolve and improve. Suno is now on Android and recently released Version 4, which boast very high-quality AI music generation.
Productivity, Research and Analysis
Perplexity opened eyes to the possibility of AI research that went beyond search. The good news is they pioneered a ‘research answer’ that went beyond web search. The other good news for us, but maybe harder on Perplexity, is that capability has been adopted by others. You can see the “web search with AI research” in Copilot, Gemini, ChatGPT, and Grok as well as Perplexity now.
Google has taken this to the next level with Gemini Deep Research. It has glimpses of AI agents doing research tasks for us.
Google also gave us NotebookLM, which went viral over its Podcast feature, but NotebookLM can also be considered an AI-based knowledge repackaging and learning tool. For example, a new NotebookLM feature lets you “join” Audio Overviews to speak to the hosts. Imagine using it to analyze a detailed report or a legal contract you uploaded, where you can ask specific questions. This has many powerful niche applications.
AI Presents and AI Futures
The past two years have been a flood of new AI models, research advances and tool improvements. There are also signs that AI is starting to mature, as AI models and applications move from novel prototypes to useful tools. As the rough edges on AI features and tools are smoothed out, we can reap the benefits of these advances by incorporating what makes most sense into our work use and personal routines.
I hope these AI models, apps, and tools help make your 2025 more productive and enjoyable, as we look forward to another year of AI improving rapidly.