Code Llama 70B: Complete Benchmark Performance Guide

meta code-llama-70b 2024-01-29 Llama 2 Community License

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

Key Specifications

Vendormeta
Versioncode-llama-70b
Release Date2024-01-29
Context Window16000 tokens
Input Modalitiestext
Output Modalitiestext
LicenseLlama 2 Community License
Documentationhttps://llama.meta.com/docs/

Benchmark Performance

Benchmark Score Unit Evaluated At Notes Source
MMLU 62.1 % 2024-01-29 5-shot view
HUMANEVAL 65.6 pass@1 2024-01-29 — view
GSM8K 61 % 2024-01-29 0-shot CoT view
MATH 34 % 2024-01-29 0-shot CoT view
BBH 72.1 % 2024-01-29 3-shot CoT view
GPQA 25.3 % 2024-01-29 0-shot view
IFEVAL 63.8 % 2024-01-29 prompt_strict view
ARC 89.8 % 2024-01-29 challenge view
MUSR 40.1 % 2024-01-29 0-shot view
WINOGRANDE 75.5 % 2024-01-29 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://ai.meta.com/blog/ · as of 2024-01-29

Compliance

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

Code Llama 70B

Model Overview

Meta Code Llama 70B 代码专用开源模型, 16K 上下文, 基于 Llama 2 微调, 支持多语言代码生成。

Core Specifications

Vendor Version Release Date Context Window Input Modalities Output Modalities License
Meta code-llama-70b 2024-01-29 16K text text Llama 2 Community License

Benchmark Performance

Benchmark Score Unit Notes
MMLU (Massive Multitask Language Understanding) 62.1 % 5-shot
HumanEval 65.6 pass@1 —
GSM8K (Grade School Math 8K) 61.0 % 0-shot CoT
MATH 34.0 % 0-shot CoT
BBH (BIG-Bench Hard) 72.1 % 3-shot CoT
GPQA 25.3 % 0-shot
IFEval 63.8 % prompt_strict
ARC 89.8 % challenge
MUSR 40.1 % 0-shot
WinoGrande 75.5 % 0-shot

Pricing

Input Output Cache Read Cache Write
— — — —

per million tokens

Strengths

  • Reliable general-purpose model.

Weaknesses

  • Proprietary, not self-hostable.
  • Context window 16K is limited.

Use Cases

  • Code generation and debugging

References