Llama 3

2024-04-17 · Source: Ollama Blog · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering · Depth: Intermediate, quick

Summary

Meta's Llama 3 large language model is now available for local execution via Ollama, marking a significant advancement over its predecessor, Llama 2. Released on April 18, 2024, Llama 3 was trained on a dataset seven times larger than Llama 2 and features a doubled context length of 8K tokens. It utilizes a 128K token vocabulary for more efficient language encoding and exhibits less than one-third of the false "refusals" compared to Llama 2. The initial release includes two parameter sizes: 8B and 70B, both accessible through simple `ollama run` commands. Future plans involve a 400B parameter model and enhanced capabilities like multimodality, multi-language support, and even longer context windows.

Key takeaway

For AI Engineers and Machine Learning Engineers looking to experiment with advanced open models locally, Llama 3's availability through Ollama offers a compelling option. You should consider integrating the 8B or 70B parameter models into your development workflows, especially for tasks requiring improved context handling and reduced refusals, and prepare for future multimodal and multilingual enhancements.

Key insights

Llama 3 offers substantial improvements over Llama 2, now locally runnable via Ollama.

Principles

• Larger datasets improve model capabilities.
• Increased token vocabulary enhances language encoding efficiency.

Method

Llama 3 can be run locally by downloading Ollama and executing `ollama run llama3` or specific parameter versions like `ollama run llama3:8b`.

In practice

• Integrate Llama 3 with LangChain via `Ollama(model="llama3")`.
• Use Llama 3 with LlamaIndex through `Ollama(model="llama3")`.

Topics

• Llama 3
• Ollama Integration
• LLM Performance
• Model Scaling
• Multimodal AI

Best for: AI Engineer, Machine Learning Engineer, Software Engineer

Related on AIssential

• Llama 3.2 goes small and multimodal — Llama 3.2 offers small, text-only models for edge devices and larger multimodal models for image reasoning.
• Tool support — Ollama now enables LLMs to interact with external tools for complex tasks via a new API feature.
• Llama 3 on Your Local Computer | Free GPT-4 Alternative — Llama 3 achieves significant performance gains through extensive training data, an improved tokenizer, and advanced fine-tuning.
• Meta Llama 3.1 405B May Spark a Tectonic Shift in the LLM Market Adoption — Meta's Llama 3.1 405B is the first open model to achieve frontier-level performance, rivaling top proprietary LLMs.
• Ollama: Unleash the Power of Local LLMs on Your Machine — Ollama enables private, cost-effective, and low-latency local LLM inference and customization via a Python library.
• 100% Free Claude Code | Run Claude Code with Local LLM with Ollama and Qwen 3.5 — Local LLMs can be integrated with Quad Code via Ollama for specific coding tasks, offering a viable alternative to API-based solutions.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Ollama Blog.