Gemma 7B: Complete Benchmark Performance Guide

google gemma-7b 2024-02-21 Gemma License

In-depth analysis of Gemma 7B performance on MMLU, HumanEval, GSM8K, MATH, BBH benchmarks, including pricing and use cases.

Key Specifications

Vendorgoogle
Versiongemma-7b
Release Date2024-02-21
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 69.6 % 2024-02-21 5-shot view
HUMANEVAL 61 pass@1 2024-02-21 — view
GSM8K 63.6 % 2024-02-21 0-shot CoT view
MATH 25.7 % 2024-02-21 0-shot CoT view
BBH 65.7 % 2024-02-21 3-shot CoT view
GPQA 37.1 % 2024-02-21 0-shot view
IFEVAL 66.9 % 2024-02-21 prompt_strict view
ARC 88.4 % 2024-02-21 challenge view
MUSR 36.2 % 2024-02-21 0-shot view
WINOGRANDE 74.1 % 2024-02-21 0-shot view

Pricing

Tier Price Currency
Input$0.22 / MtokUSD
Output$0.22 / MtokUSD
Cache Read$0 / MtokUSD
Cache Write$0 / MtokUSD

Source: https://ai.google.dev/pricing · as of 2024-02-21

Compliance

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

Gemma 7B

Model Overview

Google Gemma 7B 开源模型, 8K 上下文, 基于 Gemini 技术, 在 7B 规模上达到 Llama 2 13B 性能。

Core Specifications

Vendor Version Release Date Context Window Input Modalities Output Modalities License
Google gemma-7b 2024-02-21 8K text text Gemma License

Benchmark Performance

Benchmark Score Unit Notes
MMLU (Massive Multitask Language Understanding) 69.6 % 5-shot
HumanEval 61.0 pass@1 —
GSM8K (Grade School Math 8K) 63.6 % 0-shot CoT
MATH 25.7 % 0-shot CoT
BBH (BIG-Bench Hard) 65.7 % 3-shot CoT
GPQA 37.1 % 0-shot
IFEval 66.9 % prompt_strict
ARC 88.4 % challenge
MUSR 36.2 % 0-shot
WinoGrande 74.1 % 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

  • General chat and Q&A

References