Starting today, open source is leading the way. Introducing Llama 3.1: Our most capable models yet. Today we’re releasing a collection of new models including our long awaited 405B. Llama 3.1 delivers stronger reasoning, a larger 128K context window & improved support for 8 languages including English — among other improvements. Details in the full announcement ➡️ https://go.fb.me/hvuqhb Download the models ➡️ https://go.fb.me/11ffl7 We evaluated performance across 150+ benchmark datasets across a range of languages — in addition to extensive human evaluations in real-world scenarios. Trained on >16K NVIDIA H100 GPUs, Llama 3.1 405B is the industry leading open source foundation model and delivers state-of-the-art capabilities that rival the best closed source models in general knowledge, steerability, math, tool use and multilingual translation. We’ve also updated our license to allow developers to use the outputs from Llama models — including the 405B — to improve other models for the first time. We’re excited about how synthetic data generation and model distillation workflows with Llama will help to advance the state of AI. As Mark Zuckerberg shared this morning, we have a strong belief that open source will ensure that more people around the world have access to the benefits and opportunities of AI and that’s why we continue to take steps on the path for open source AI to become the industry standard. With these releases we’re setting the stage for unprecedented new opportunities and we can’t wait to see the innovation our newest Llama models will unlock across all levels of the AI community.
AI at Meta
Research Services
Menlo Park, California 854,727 followers
Together with the AI community, we’re pushing boundaries through open science to create a more connected world.
About us
Through open science and collaboration with the AI community, we are pushing the boundaries of artificial intelligence to create a more connected world. We can’t advance the progress of AI alone, so we actively engage with the AI research and academic communities. Our goal is to advance AI in Infrastructure, Natural Language Processing, Generative AI, Vision, Human-Computer Interaction and many other areas of AI enable the community to build safe and responsible solutions to address some of the world’s greatest challenges.
- Website
-
https://ai.meta.com/
External link for AI at Meta
- Industry
- Research Services
- Company size
- 10,001+ employees
- Headquarters
- Menlo Park, California
- Specialties
- research, engineering, development, software development, artificial intelligence, machine learning, machine intelligence, deep learning, computer vision, engineering, computer vision, speech recognition, and natural language processing
Updates
-
Announced today, NVIDIA NIM Agent Blueprints include everything enterprises need to build and deploy custom apps using Llama 3.1 405B. • Sample applications • Reference code • Customization documentation • A Helm chart for deployment More NIM Agent Blueprints are on the way covering use cases like content generation, software engineering, retail shopping advising and more.
📣 Announcing NVIDIA NIM Agent Blueprints, reference applications for every enterprise to make their own #AI. ➡️ https://nvda.ws/3z1HTUO These new NIM Agent Blueprints equip millions of enterprise developers with reference applications for building and deploying generative AI use cases such as digital humans, multimodal PDF data extraction, and virtual screening.
-
-
Announced this morning and verified by Artificial Analysis, Cerebras Systems Inference is capable of serving Llama 3.1 70B at 450 tokens/sec and Llama 3.1 8B at 1,850 tokens/sec! This order of magnitude increase in inference speed for Llama 3.1 could unlock all new types of use cases for the developer community and enterprises.
Meet Cerebras Inference – the fastest inference for generative AI! 🏎️ Speed: 1,800 tokens/sec for Llama 3.1-8B and 450 tokens/sec for Llama 3.1-70B, 20x faster than NVIDIA GPU-based hyperscale clouds. 💸 Price: Cerebras Inference offers the industry’s best price-performance at 10c per million tokens for Llama 3.1-8B and 60c per million tokens for Llama-3.1 70B. 🎯 Accuracy: Cerebras Inference uses native 16-bit weights for all models, ensuring the highest accuracy responses. 🔓 Access: Cerebras Inference is open to everyone today via chat and API access. All powered by our third-generation Wafer Scale Engine (WSE-3). Try it now 👉 https://lnkd.in/gEJJ2pfY Press Release: https://lnkd.in/gtF5fxHt Blog: https://lnkd.in/gZ46q4cD
-
As part of the release of Llama 3.1, we also released new trust and safety artifacts including CyberSecEval 3. We've published a new research paper on this work to continue the conversation with researchers and developers on empirically measuring LLM cybersecurity risks and capabilities. CyberSecEval 3 research paper ➡️ https://go.fb.me/yv32a9
-
-
New course on DeepLearning.AI: Improving Accuracy of LLM Applications. More details ➡️ https://go.fb.me/zfwvd8 Created in collaboration with DeepLearning.AI, Meta and Lamini, this free course covers: • Building a text to SQL agent with Llama 3. • Evaluation frameworks. • Instruction & memory fine-tuning. • Breaking fine-tuning myths. • PEFT techniques like LoRA and MoME. • Processes for generating training data. • More
-
-
New research from Meta FAIR: Unibench is a unified implementation of 50+ VLM benchmarks spanning a comprehensive range of carefully categorized capabilities from object recognition to spatial awareness, counting and much more. Research paper ➡️ https://go.fb.me/fa97z9 UniBench repo ➡️ https://go.fb.me/dxxz4w
-
-
🆕 Research paper from GenAI at Meta: Imagine yourself: Tuning-Free Personalized Image Generation. Research paper ➡️ https://go.fb.me/h6aa5x This paper introduces a state-of-the-art model designed for personal image generation. Want to try it? The feature is available now as a beta in Meta AI for users in the US.
-
-
-
-
-
+1
-
-
AI at Meta reposted this
We have a unique and technically-driven culture in GenAI and AI at Meta that enables our researchers and engineers to drive big ideas that push the state of AI as a technology forward — but also put those ideas in practice to enable impact in products at scale. This new post shares a little bit of how this culture shaped Llama 3.1. I’m proud of what this team has been able to do in such a short time, and the culture we have built to enable this: https://lnkd.in/gSafPKKA Does driving both academic progress and products at scale excite you? We’re hiring across roles Research Science, SWE and ML Eng roles in GenAI: https://lnkd.in/gapZN9Kq
-
Using structured weight pruning and knowledge distillation, the NVIDIA research team refined Llama 3.1 8B into a new Llama-3.1-Minitron 4B. They're releasing the new models on Hugging Face and shared a deep dive on their approach ➡️ https://go.fb.me/8khfyr
-
-
Five papers to add to your reading list from AI researchers at Meta at #ACL2024NLP. 🔗 Research Papers 1. MuTox: Universal MUltilingual Audio-based TOXicity Dataset and Zero-shot Detector: https://go.fb.me/y7vrjg 2. Are Female Carpenters like Blue Bananas? A Corpus Investigation of Occupation Gender Typicality: https://go.fb.me/cdfzig 3. LM Transparency Tool: Interactive Tool for Analyzing Transformer Language Models: https://go.fb.me/5ngnnp 4. LayerSkip: Enabling Early Exit Inference and Self-Speculative Decoding: https://go.fb.me/96bbg1 5. The Belebele Benchmark: a Parallel Reading Comprehension Dataset in 122 Language Variants: https://go.fb.me/irdq2z
-