Compare Llama 3.1 405B Instruct vs. GPT-4 32K 0613

Pricing About Us Blog Docs

Compare

Interested in comparing the performance of these models in production? Get a demo of Context.ai

Overview

Llama 3.1 405B Instruct was released about 1 year after GPT-4 32K 0613.

	Llama 3.1 405B Instruct	GPT-4 32K 0613
Provider The entity that provides this model.	Meta	OpenAI
Input Context Window The number of tokens supported by the input context window.	128K tokens	32.8K tokens
Maximum Output Tokens The number of tokens that can be generated by the model in a single request.	2,048 tokens	Not specified.
Release Date When the model was first released.	July 23rd, 2024 about 2 months ago 2024-07-23	June 13th, 2023 over 1 year ago 2023-06-13
Knowledge Cutoff Limit on the knowledge base used by the model.	December 2023 December 2023	Unknown
Open Source	Yes	No
API Providers The providers that offer this model. (This is not an exhaustive list.)	Azure AI, AWS Bedrock, Google Cloud Vertex AI Model Garden, NVIDIA NIM, IBM watsonx	OpenAI

Pricing

	Llama 3.1 405B Instruct	GPT-4 32K 0613
Input Cost of input data provided to the model.	Pricing not available.	$60.00 per million tokens
Output Cost of output tokens generated by the model.	Pricing not available.	$120.00 per million tokens

Benchmarks

Compare relevant benchmarks between Llama 3.1 405B Instruct and GPT-4 32K 0613.

	Llama 3.1 405B Instruct	GPT-4 32K 0613
MMLU Evaluating LLM knowledge acquisition in zero-shot and few-shot settings.	85.2 (5-shot) Source	Benchmark not available.
MMMU A wide ranging multi-discipline and multimodal benchmark.	Benchmark not available.	Benchmark not available.
HellaSwag A challenging sentence completion benchmark.	Benchmark not available.	Benchmark not available.
GSM8K Grade-school math problems benchmark.	96.8 (8-shot) Source	Benchmark not available.
HumanEval A benchmark to measure functional correctness for synthesizing programs from docstrings.	Benchmark not available.	Benchmark not available.
MATH Benchmark performance on Math problems ranging across 5 levels of difficulty and 7 sub-disciplines.	73.8 (0-shot) Source	Benchmark not available.