GenAIWiki

Llama 3.1 405B Instruct

CurrentLatest

Meta’s largest open-weights instruct checkpoint in the Llama 3.1 family, aimed at strong reasoning and coding quality with a permissive license for research and customization.

Provider

Meta

Model family

Meta Llama

Open weights LLM

Cost tier

Open / entry

Status

Current

Release Jul 23, 2024

Why teams choose it

🧠

Serving 405B-class weights requires multi-GPU inference—budget hardware and ops explicitly.

Serving 405B-class weights requires multi-GPU inference—budget hardware and ops explicitly.

📎

License obligations include acceptable use and attribution—legal review before redistribution.

License obligations include acceptable use and attribution—legal review before redistribution.

Tradeoffs to know

  • Not a drop-in for tiny edge devices—needs serious infrastructure.
  • Safety tooling is operator responsibility when self-hosting.

When not to use this

  • Self-hosting outcomes depend on hardware, quantization, and ops maturity—budget time beyond swapping an API hostname.
  • May demand more instrumentation than SaaS-managed APIs to duplicate latency, failover, and support guarantees.
  • Benchmark prompts and regressions continuously before rewriting entire routing tables around weights.

Technical specs

Inputs
text
Outputs
text
Capabilities
reasoning, coding, fine-tuning
License
Llama 3.1 Community License
Model string
llama-3-1-405b-instruct

Benchmarks

{
  "gpqa": 51.1,
  "mmlu": 88.6
}

Meta Llama family lineup


Compare with