Gemma 2 27B: Complete Benchmark Performance Guide

google gemma-2-27b 2024-06-27 Gemma License

In-depth analysis of Gemma 2 27B performance on MMLU, HumanEval, GSM8K, MATH, BBH benchmarks, including pricing and use cases.

Key Specifications

Vendorgoogle
Versiongemma-2-27b
Release Date2024-06-27
Context Window8192 tokens
Input Modalitiestext
Output Modalitiestext
LicenseGemma License
Documentationhttps://ai.google.dev/gemma/docs

Benchmark Performance

Benchmark Score Unit Evaluated At Notes Source
MMLU 77.4 % 2024-06-27 5-shot view
HUMANEVAL 72.4 pass@1 2024-06-27 — view
GSM8K 77.4 % 2024-06-27 0-shot CoT view
MATH 35.5 % 2024-06-27 0-shot CoT view
BBH 71.2 % 2024-06-27 3-shot CoT view
GPQA 35.4 % 2024-06-27 0-shot view
IFEVAL 72.5 % 2024-06-27 prompt_strict view
ARC 93.1 % 2024-06-27 challenge view
MUSR 53 % 2024-06-27 0-shot view
WINOGRANDE 80.5 % 2024-06-27 0-shot view

Pricing

Tier Price Currency
Input$0.7 / MtokUSD
Output$0.7 / MtokUSD
Cache Read$0 / MtokUSD
Cache Write$0 / MtokUSD

Source: https://ai.google.dev/pricing · as of 2024-06-27

Compliance

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

Gemma 2 27B

Model Overview

Google Gemma 2 27B 开源模型, 8K 上下文, 在 27B 规模上接近 GPT-4 性能, 推理能力突出。

Core Specifications

Vendor Version Release Date Context Window Input Modalities Output Modalities License
Google gemma-2-27b 2024-06-27 8K text text Gemma License

Benchmark Performance

Benchmark Score Unit Notes
MMLU (Massive Multitask Language Understanding) 77.4 % 5-shot
HumanEval 72.4 pass@1 —
GSM8K (Grade School Math 8K) 77.4 % 0-shot CoT
MATH 35.5 % 0-shot CoT
BBH (BIG-Bench Hard) 71.2 % 3-shot CoT
GPQA 35.4 % 0-shot
IFEval 72.5 % prompt_strict
ARC 93.1 % challenge
MUSR 53.0 % 0-shot
WinoGrande 80.5 % 0-shot

Pricing

Input Output Cache Read Cache Write
— — — —

per million tokens

Strengths

  • Reliable general-purpose model.

Weaknesses

  • Proprietary, not self-hostable.
  • Context window 8K is limited.

Use Cases

  • Code generation and debugging

References