Meta
Llama 3.1 405B Instruct
Open weights LLM · Release Jul 23, 2024 · Llama 3.1 Community License
Large open-weights instruct model competitive on reasoning and coding benchmarks with permissive licensing for customization.
Modalities
What goes in and what comes out.
Inputs
text
Outputs
text
Capabilities
reasoning, coding, fine-tuning
Benchmarks snapshot
Structured JSON for reproducible comparisons.
{
"gpqa": 51.1,
"mmlu": 88.6
}Related on GenAIWiki
Same provider, tooling that cites the model, or prompts tuned for it.
Meta
LLaMA 3 70B
LLaMA 3 70B features 70 billion parameters and a context window of 32k tokens, optimized for high-performance text generation and understanding across diverse tasks.
Meta
LLaMA 3 8B
LLaMA 3 8B is a compact model with 8 billion parameters, designed for efficient text generation and understanding with a context window of 8k tokens.
Inference
Groq
GroqCloud offers very low-latency, high-throughput LLM inference using Groq’s LPU-style hardware, with OpenAI-compatible APIs for select open and partner models aimed at interactive and batch production workloads.
ML platform
Hugging Face
Hub for open models, datasets, and Spaces demos, plus Inference Endpoints, Transformers, and enterprise features for teams that train, fine-tune, or serve open-weight and partner models at scale.
Orchestration
LangChain
Application framework for orchestrating LLM workflows, tool calling, retrieval, and agents across multiple providers in Python and TypeScript ecosystems.
Compute
Modal
Serverless compute platform for AI inference and batch workloads, offering GPU execution, scalable workers, and code-first deployment patterns for model-powered applications.