NVIDIA Llama 3.1 Nemotron 70B: Complete Benchmark Performance Guide

other nvidia-llama-3-1-nemotron-70b 2024-10-17 Llama 3.1 Community License

In-depth analysis of NVIDIA Llama 3.1 Nemotron 70B performance on MMLU, HumanEval, GSM8K, MATH, BBH benchmarks, including pricing and use cases.

Key Specifications

Vendorother
Versionnvidia-llama-3-1-nemotron-70b
Release Date2024-10-17
Context Window131072 tokens
Input Modalitiestext
Output Modalitiestext
LicenseLlama 3.1 Community License
Documentationhttps://huggingface.co/models

Benchmark Performance

Benchmark Score Unit Evaluated At Notes Source
MMLU 80.4 % 2024-10-17 5-shot view
HUMANEVAL 66.4 pass@1 2024-10-17 — view
GSM8K 77.8 % 2024-10-17 0-shot CoT view
MATH 40.3 % 2024-10-17 0-shot CoT view
BBH 76.6 % 2024-10-17 3-shot CoT view
GPQA 37.6 % 2024-10-17 0-shot view
IFEVAL 69.6 % 2024-10-17 prompt_strict view
ARC 93.2 % 2024-10-17 challenge view
MUSR 47.7 % 2024-10-17 0-shot view
WINOGRANDE 84.7 % 2024-10-17 0-shot view

Pricing

Tier Price Currency
Input$0.9 / MtokUSD
Output$0.9 / MtokUSD
Cache Read$0 / MtokUSD
Cache Write$0 / MtokUSD

Source: https://huggingface.co/models · as of 2024-10-17

Compliance

  • Data Residency: self-host
  • SOC2: ✗
  • HIPAA: ✗
  • GDPR: ✗
  • ISO 27001: ✗

NVIDIA Llama 3.1 Nemotron 70B

Model Overview

NVIDIA Llama 3.1 Nemotron 70B, 131K 上下文, NVIDIA 优化 Llama 3.1, 在 Arena 上接近 GPT-4o。

Core Specifications

Vendor Version Release Date Context Window Input Modalities Output Modalities License
Other nvidia-llama-3-1-nemotron-70b 2024-10-17 131K text text Llama 3.1 Community License

Benchmark Performance

Benchmark Score Unit Notes
MMLU (Massive Multitask Language Understanding) 80.4 % 5-shot
HumanEval 66.4 pass@1 —
GSM8K (Grade School Math 8K) 77.8 % 0-shot CoT
MATH 40.3 % 0-shot CoT
BBH (BIG-Bench Hard) 76.6 % 3-shot CoT
GPQA 37.6 % 0-shot
IFEval 69.6 % prompt_strict
ARC 93.2 % challenge
MUSR 47.7 % 0-shot
WinoGrande 84.7 % 0-shot

Pricing

Input Output Cache Read Cache Write
— — — —

per million tokens

Strengths

  • MMLU score 80.4, strong knowledge reasoning.

Weaknesses

  • Proprietary, not self-hostable.

Use Cases

  • General chat and Q&A

References