Phi-3.5 (MoE, Mini & Vision) : The NEW BEST Small Model is finally here! (Beats Llama-3.1, Mistral)

Join this channel to get access to perks:

In this video, I'll be fully testing the New Phi-3.5 Model to check if it's really good. I'll also be trying to find out if it can really beat Llama-3.1, Claude 3.5 Sonnet, GPT-4O, DeepSeek & Qwen-2. This model is fully opensource and can be used for FREE. It is even better in Coding Tasks and is also really good at doing Text-To-Application, Text-To-Frontend and other things as well. I'll be testing it to find out if it can really beat other LLMs and i'll also be telling you that how you can use it.

-----
Key Takeaways:

🔥 Microsoft's New Phi-3.5 Models: Discover the latest AI advancements with Phi-3.5 Vision, Phi-3.5 Mini, and the powerful Mixture of Experts model, ideal for AI enthusiasts and developers.

🧠 Mixture of Experts Explained: Understand how Microsoft’s Phi-3.5 MoE model intelligently routes prompts to specialized AI experts, enhancing performance in diverse tasks, perfect for AI researchers.

⚡ Top-Performing Benchmarks: See how Phi-3.5 models outperform rivals like Llama and Gemini in key AI benchmarks, making them essential tools for AI innovation and research.

💻 Phi-3.5 Mini & Vision Capabilities: Learn how the compact 3.8B parameter Mini model and the cutting-edge Vision model are transforming AI, with powerful features in a small package, great for developers.

🚀 Running AI Locally: Explore how Phi-3.5 models, especially the MoE, are designed for efficient local performance, making AI more accessible to developers and tech enthusiasts.

🎯 Real-World Use Cases: From solving complex math problems to generating code, see how Phi-3.5 models handle diverse tasks, proving their versatility in AI applications.

🔍 Future of AI with Microsoft: Get a glimpse into the future of AI as Microsoft pushes the boundaries with their latest Phi-3.5 models, essential viewing for anyone interested in cutting-edge AI technology.

------
Timestamps:

00:00 - Introduction
00:08 - About Phi-3.5 Mini (MoE, Mini & Vision)
05:07 - Testing
11:16 - Conclusion & Ending

[jackflash]

Seems like we can't go 1 week without something spectacular being introduced.
What a time to be alive.
Thanks for the great breakdown and testing. The MOE is quite impressive. Ollama needs to hurry and make that compatible.

[ym-xxkj]

2nd!

lets GO, can't wait for things to advance more

[detectivelon]

Great video thanks for the explanations.

[vlee]

I don't know why they are comparing Phi-3.5 to smaller models though... They should use some medium models like gemma 2 27b.

The architecture is different and uses only 6.6b active parameters at a time but it's still taking a substantial amount of memory.

[SansaWood-se]

That helps a lot! Thank you. I would like to have a little talk with you, and if it is possible for me to reach out?

[daryladhityahenry]

Hi AICodeKing!! Can you take a look at Falcon Mamba? I already wait for model like that for a looong time since mamba news.. But right now, I don't know how to run it. But it seems pretty cool, I really want to know if possible :D:D. And if it's really working well, I don't think transformers model can compete with mamba because of context length.

[MeinDeutschkurs]

These models are strange.I cannot say anything. Have to wait for ollama support.

[QHawk]

Can I have like my local files and environment on my local PC, but use colab, lightningAI or kaggle or other services GPU via terminal? I use Linux for AI stuff . say to make a chatgpt style chatbot with falcon 7b/40b model.

[aivy-aigeneratedmusic]

Your questions are almost only math and coding. Maybe mix it up more with new questions ?