Mistral Nemo: Complete Benchmark Performance Guide

mistral nemo 2024-07-18 Apache 2.0

In-depth analysis of Mistral Nemo performance on MMLU, HumanEval, GSM8K, MATH, BBH benchmarks, including pricing and use cases.

Key Specifications

Vendormistral
Versionnemo
Release Date2024-07-18
Context Window128000 tokens
Input Modalitiestext
Output Modalitiestext
LicenseApache 2.0
Documentationhttps://docs.mistral.ai/

Benchmark Performance

Benchmark Score Unit Evaluated At Notes Source
MMLU 69.2 % 2024-07-18 5-shot view
HUMANEVAL 69.4 pass@1 2024-07-18 — view
GSM8K 70 % 2024-07-18 0-shot CoT view
MATH 43.6 % 2024-07-18 0-shot CoT view
BBH 71.2 % 2024-07-18 3-shot CoT view
GPQA 36.1 % 2024-07-18 0-shot view
IFEVAL 56 % 2024-07-18 prompt_strict view
ARC 90.5 % 2024-07-18 challenge view
MUSR 38.5 % 2024-07-18 0-shot view
WINOGRANDE 73.2 % 2024-07-18 0-shot view

Pricing

Tier Price Currency
Input$0.18 / MtokUSD
Output$0.18 / MtokUSD
Cache Read$0 / MtokUSD
Cache Write$0 / MtokUSD

Source: https://mistral.ai/technology/ · as of 2024-07-18

Compliance

  • Data Residency: EU
  • SOC2: ✓
  • HIPAA: ✗
  • GDPR: ✓
  • ISO 27001: ✓

Mistral Nemo

Model Overview

Mistral Nemo 12B 开源模型, 128K 上下文, 与 NVIDIA 合作开发, 适合本地部署与多语言任务。

Core Specifications

Vendor Version Release Date Context Window Input Modalities Output Modalities License
Mistral nemo 2024-07-18 128K text text Apache 2.0

Benchmark Performance

Benchmark Score Unit Notes
MMLU (Massive Multitask Language Understanding) 69.2 % 5-shot
HumanEval 69.4 pass@1 —
GSM8K (Grade School Math 8K) 70.0 % 0-shot CoT
MATH 43.6 % 0-shot CoT
BBH (BIG-Bench Hard) 71.2 % 3-shot CoT
GPQA 36.1 % 0-shot
IFEval 56.0 % prompt_strict
ARC 90.5 % challenge
MUSR 38.5 % 0-shot
WinoGrande 73.2 % 0-shot

Pricing

Input Output Cache Read Cache Write
— — — —

per million tokens

Strengths

  • Context window 128K.

Weaknesses

  • Proprietary, not self-hostable.

Use Cases

  • Long document summarization

References