We’ve just made some huge additions to Magai in this release. Here are the highlights:
- Grok 2 – the LLM with personality
- Nemotron – GPT-4o at 1/30th the cost
- Luma Dream Machine – text-to-video!
- Runway – best in class image-to-video!
This release actually contains over 30 different improvements and updates. I’ll cover more of these at the end, but wanted to focus on the most impactful updates right away.
Let’s dive in!
Grok 2 – the LLM with personality
The flagship AI model from xAI has only been accessible to X (formerly Twitter) Premium subscribers for a while. Only recently did xAI allow 3rd parties to access their LLM outside of the X app. And now Magai users can experience all of Grok 2’s quirkiness right inside their favorite AI platform.
According to benchmark studies it is said that Grok 2 has near-GPT-4o capabilities. In my testing it performs exceedingly well at day-to-day tasks. The big difference is the “personality” it seems to have. It typically mirrors the users personality based on prompts given, and has just a hint of snark.
If you like your AI to have some more “spunk” in it without having to build a Persona for it, this might be the model you’ve been waiting for.
Nemotron – GPT-4o at 1/30th the Cost
OpenAI just got a HUGE competitor this week when Nvidia (the computer chip manufacturer who has basically built all of the GPUs responsible for training all the worlds best LLMs) released their own flagship model, officially named Llama 3.1 Nemotron 70B Instruct. (Just rolls right off the tongue, doesn’t it?)
We’ll stick to just calling it Nemotron. And it’s now available to all Magai users.
According to benchmark studies, Nemotron scores higher than GPT-4o in many areas. And the craziest part is that it runs at a fraction of the cost.
In my limited testing so far, it 100% delivers on the promises. There are some small things that GPT-4o seems to be better at, but I haven’t had enough time to rule out flukes and get a good baseline.
Give it a try and see what you think!
Luma Dream Machine – our first text-to-video model!
We’ve been waiting to introduce the first text-to-video model for Magai users, and the time has finally arrived!
Just like an image generator, you describe the video you want with text, click the “Generate” button, and Luma Dream will produce your described video.
Pretty magical.
Runway – Best-in-class image-to-video!
Runway has been praised as the best generative video product on the market. It can take an existing image and turn it into a 5-10 second long video with stunning accuracy. This is very much like our Leonardo Motion integration, but the quality is noticeably better.
My recommendation is to use either Flux 1 Pro or Flux 1.1 Pro if your looking for hyper realistic results.
More updates worth noting
As I said before, there were a total of over 30 updates, improvements, and fixes packed into this latest update. This is on top of a few updates before that which also included dozens and dozens of improvements. Here are some of the most noteworthy:
- Flux 1.1 Pro added
- Flux option for Inpainting
- Flux Paramaters (such as Inference Steps, Guidance Scale, and Safety Tolerance) added where applicable
- Leonardo Upscaler added to Chat image editor
- Export Chat to PDF
- Remove character limit on Workspace Custom Context
- AI Auto-Title improvements
- Massive UI/UX improvements to Editor page (especially for Mobile)
- Web Search function captures more results
- High Context Multiplier thresholds loosened up
- Added “Re-run” Generation in Editor to easily re-apply the same settings for a new image generation
If it’s not evident, we’ve been working like crazy to continually improve the app in every single way. Over the last 3 months our team has closed over 200 development tickets, ensuring that Magai is the best it can possibly be for every single user.
And this is the mission we will continue, working as hard as we can to make sure you have the most powerful AI tools all in one place.
Now back to work. ☕️
