Meta: Llama 3.1 405B (base)

Meta: Llama 3.1 405B (base)

meta-llama · Released Aug 2, 2024 Professional
Intelligence #10 / 576
82.0 Our Score
Speed #224 / 268
48.8 tokens / sec
Input #533 / 576
$4.00 per 1M tokens
Output #448 / 576
$4.00 per 1M tokens
Context #393 / 576
32,768 tokens

Analysis Summary

Meta's Llama 3.1 405B base is the untuned foundation model from the Llama 3.1 family. With an intelligence index of 17.4 and MMLU-Pro of 0.732, it shows reasonable general knowledge, but as a base model it lacks instruction tuning, making it unsuitable for direct deployment in business workflows without further fine-tuning. It also has no tool use or function calling support, and the 32K context window is a constraint for long-document tasks.

For businesses, this model is primarily relevant as a fine-tuning base for teams with the infrastructure and expertise to adapt it for specific use cases. It is not a drop-in solution for content generation, coding assistance, or agentic workflows in its current form.

At $4.00 per million tokens in and out, it is expensive for a base model, especially given the additional investment required to make it production-ready. Only teams with clear fine-tuning roadmaps and the capacity to run large models should consider it.

Assessed June 6, 2026

Editorial notes

Llama 3.1 405B base from Meta is a large open-weight model with reasonable reasoning benchmarks, but its base (non-instruct) nature, small 32K context, and high pricing limit its practical business utility.

Rankings consider pricing, capabilities, benchmarks, and real-world applicability and are refreshed as new models launch. Feedback?

Performance Profile

Intelligence3Technical2Value6Content3.6
Intelligence 3/10
Technical 2/10
Content 3.6/10
Value 6/10

How Meta: Llama 3.1 405B (base) compares

Meta: Llama 3.1 405B (base) ranks #199 of 377 AI models we track for overall intelligence, #187 of 314 for coding, #237 of 289 for agentic tasks. Its 33K-token context window is larger than 32% of the models we list. At $4.00 per million input tokens it is cheaper than 7% of comparable models.

About Meta: Llama 3.1 405B (base)

Meta's latest class of model (Llama 3.1) launched with a variety of sizes & flavors. This is the base 405B pre-trained version. It has demonstrated strong performance compared to leading closed-source models in human evaluations. Usage of this model is subject to Meta's Acceptable Use Policy.

405B Parameters

Architecture Detail

Instruct Typenone

Performance Indices

Source: Artificial Analysis

17.4 Intelligence Index
14.5 Coding Index
12.9 Agentic Index
3 Math Index

Benchmark Scores

Intelligence

GPQA Diamond 51.5% Graduate-level scientific reasoning
HLE 4.2% Humanity's Last Exam
MMLU Pro 73.2% Multi-task language understanding
MATH 500 70.3% Mathematical problem-solving
AIME 21.3% Competition mathematics
AIME 2025 3% Competition mathematics (2025)
SciCode 29.9% Scientific computing

Technical

LiveCodeBench 30.5% Live coding evaluation
TerminalBench Hard 6.8% Agentic terminal tasks
τ²-Bench 19% Conversational agent benchmark

Content

IFBench 39% Instruction following
LCR 24.3% Long-context reasoning

Benchmark data from Artificial Analysis and Hugging Face

How does Meta: Llama 3.1 405B (base) stack up?

Compare side-by-side with other professional models.

Compare Models

Model Information

OpenRouter ID meta-llama/llama-3.1-405b
Providermeta-llama
Model FamilyLlama 3
Release Date August 2, 2024
Context Length32,768 tokens
Max Completion32,768 tokens
Status Active

Pricing

Token Type Cost per 1M tokens Cost per 1K tokens
Input $4.00 $0.004000
Output $4.00 $0.004000

Leaderboard Categories

Frequently asked questions about Meta: Llama 3.1 405B (base)

How much does Meta: Llama 3.1 405B (base) cost?

Meta: Llama 3.1 405B (base) costs $4.00 per million input tokens and $4.00 per million output tokens.

What is the context window of Meta: Llama 3.1 405B (base)?

Meta: Llama 3.1 405B (base) has a context window of 32,768 tokens (33K).

Is Meta: Llama 3.1 405B (base) good for coding?

On our coding benchmark index, Meta: Llama 3.1 405B (base) ranks #187 of 314 models, placing it in the broader range of the field for code generation and debugging.

Who created Meta: Llama 3.1 405B (base)?

Meta: Llama 3.1 405B (base) is developed by Meta and was released on August 2, 2024.