Post
597
Mert Erbak PRO
merterbak
AI & ML interests
NLP and Image Processing
Recent Activity
updated
a Space
about 5 hours ago
merterbak/grok
updated
a Space
about 5 hours ago
merterbak/gemma-3
updated
a Space
about 5 hours ago
merterbak/DeepSeek-R1-Distill-Qwen-1.5B
Organizations
Posts
4
Post
2852
Meta has unveiled its Llama 4 🦙 family of models, featuring native multimodality and mixture-of-experts architecture. Two model families are available now:
Models🤗: meta-llama/llama-4-67f0c30d9fe03840bc9d0164
Blog Post: https://ai.meta.com/blog/llama-4-multimodal-intelligence/
HF's Blog Post: https://huggingface.co/blog/llama4-release
- 🧠 Native Multimodality - Process text and images in a unified architecture
- 🔍 Mixture-of-Experts - First Llama models using MoE for incredible efficiency
- 📏 Super Long Context - Up to 10M tokens
- 🌐 Multilingual Power - Trained on 200 languages with 10x more multilingual tokens than Llama 3 (including over 100 languages with over 1 billion tokens each)
🔹 Llama 4 Scout
- 17B active parameters (109B total)
- 16 experts architecture
- 10M context window
- Fits on a single H100 GPU
- Beats Gemma 3, Gemini 2.0 Flash-Lite, and Mistral 3.1
🔹 Llama 4 Maverick
- 17B active parameters (400B total)
- 128 experts architecture
- It can fit perfectly on DGX H100(8x H100)
- 1M context window
- Outperforms GPT-4o and Gemini 2.0 Flash
- ELO score of 1417 on LMArena currently second best model on arena
🔹 Llama 4 Behemoth (Coming Soon)
- 288B active parameters (2T total)
- 16 experts architecture
- Teacher model for Scout and Maverick
- Outperforms GPT-4.5, Claude Sonnet 3.7, and Gemini 2.0 Pro on STEM benchmarks
Models🤗: meta-llama/llama-4-67f0c30d9fe03840bc9d0164
Blog Post: https://ai.meta.com/blog/llama-4-multimodal-intelligence/
HF's Blog Post: https://huggingface.co/blog/llama4-release
- 🧠 Native Multimodality - Process text and images in a unified architecture
- 🔍 Mixture-of-Experts - First Llama models using MoE for incredible efficiency
- 📏 Super Long Context - Up to 10M tokens
- 🌐 Multilingual Power - Trained on 200 languages with 10x more multilingual tokens than Llama 3 (including over 100 languages with over 1 billion tokens each)
🔹 Llama 4 Scout
- 17B active parameters (109B total)
- 16 experts architecture
- 10M context window
- Fits on a single H100 GPU
- Beats Gemma 3, Gemini 2.0 Flash-Lite, and Mistral 3.1
🔹 Llama 4 Maverick
- 17B active parameters (400B total)
- 128 experts architecture
- It can fit perfectly on DGX H100(8x H100)
- 1M context window
- Outperforms GPT-4o and Gemini 2.0 Flash
- ELO score of 1417 on LMArena currently second best model on arena
🔹 Llama 4 Behemoth (Coming Soon)
- 288B active parameters (2T total)
- 16 experts architecture
- Teacher model for Scout and Maverick
- Outperforms GPT-4.5, Claude Sonnet 3.7, and Gemini 2.0 Pro on STEM benchmarks
Collections
1
-
Attention Is All You Need
Paper • 1706.03762 • Published • 59 -
LoRA Learns Less and Forgets Less
Paper • 2405.09673 • Published • 89 -
DeepSeek LLM: Scaling Open-Source Language Models with Longtermism
Paper • 2401.02954 • Published • 47 -
RAFT: Adapting Language Model to Domain Specific RAG
Paper • 2403.10131 • Published • 73
spaces
7
pinned
Running
on
Zero
7
Gemma 3
💎
Chat with multimodal gemma-3-12b-it or gemma-3-4b-it models
pinned
Running
on
Zero
10
Phi 4
🐨
Chat with Microsoft's phi-4 or phi-4-mini models
Running
11
Grok Chatbot
🚀
Try XAI's Grok 2 vision model
Running
1
DeepSeek R1 Distill Qwen 1.5B
💬
Try DeepSeek-R1-Distill-Qwen-1.5B on CPU
Running
17
Mistral OCR
🌆
Try out Mistral's latest OCR with pdfs and images
Running
on
Zero
3
RAG Llama
📖
RAG Chatbot with Llama3.1-8B for PDFs