Gemma 2 9B: Complete Benchmark Performance Guide

google gemma-2-9b 2024-06-27 Gemma License

In-depth analysis of Gemma 2 9B performance on MMLU, HumanEval, GSM8K, MATH, BBH benchmarks, including pricing and use cases.

Key Specifications

Vendorgoogle
Versiongemma-2-9b
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 64.3 % 2024-06-27 5-shot view
HUMANEVAL 60.9 pass@1 2024-06-27 — view
GSM8K 64.3 % 2024-06-27 0-shot CoT view
MATH 28.9 % 2024-06-27 0-shot CoT view
BBH 66.7 % 2024-06-27 3-shot CoT view
GPQA 30.3 % 2024-06-27 0-shot view
IFEVAL 64.3 % 2024-06-27 prompt_strict view
ARC 90.3 % 2024-06-27 challenge view
MUSR 44.1 % 2024-06-27 0-shot view
WINOGRANDE 72.5 % 2024-06-27 0-shot view

Pricing

Tier Price Currency
Input$0.3 / MtokUSD
Output$0.3 / 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 9B

Model Overview

Google Gemma 2 9B 开源模型, 8K 上下文, 在 9B 规模上超越 Llama 3 8B, 推理能力接近 70B 模型。

Core Specifications

Vendor Version Release Date Context Window Input Modalities Output Modalities License
Google gemma-2-9b 2024-06-27 8K text text Gemma License

Benchmark Performance

Benchmark Score Unit Notes
MMLU (Massive Multitask Language Understanding) 64.3 % 5-shot
HumanEval 60.9 pass@1 —
GSM8K (Grade School Math 8K) 64.3 % 0-shot CoT
MATH 28.9 % 0-shot CoT
BBH (BIG-Bench Hard) 66.7 % 3-shot CoT
GPQA 30.3 % 0-shot
IFEval 64.3 % prompt_strict
ARC 90.3 % challenge
MUSR 44.1 % 0-shot
WinoGrande 72.5 % 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

  • General chat and Q&A

References