LightweightOpen SourceGoogle

Gemma 2 27B

Gemma 2 27B is Google's largest open-source lightweight model in the Gemma family, offering 27 billion parameters with an 8K token context window.

Context 8K
Tier Lightweight
Knowledge Feb 2024
License Open Source
Input from
$0.650 / 1M tokens
across 2 providers
Compare PricesModel Page →Paper

API Pricing

Cheapest on OpenRouter — 10% below avg
ProviderInput / 1MOutput / 1MUpdated
OpenRouter logo
OpenRouter
$0.650$0.6504/14/2026
Together AI logo
Together AI
$0.800$0.8004/14/2026

Prices updated daily. Last check: 4/14/2026

Model Details

General

Creator
Google
Family
Gemma
Tier
Lightweight
Context Window
8K
Knowledge Cutoff
Feb 2024
Modalities
Text

Capabilities

Tool Calling
No
Open Source
Yes
Subtypes
Chat Completion

Strengths & Limitations

  • Open-source with downloadable model weights for local deployment
  • 27 billion parameters provide substantial capability within lightweight tier
  • No API dependencies required for inference
  • Supports fine-tuning and model customization
  • February 2024 knowledge cutoff offers relatively recent training data
  • Chat completion format optimized for conversational interactions
  • Can run on consumer hardware with appropriate specifications
  • No tool calling or function execution capabilities
  • Text-only modality without image or multimodal support
  • 8K context window smaller than many competing models
  • Lightweight tier positioning limits complex reasoning capabilities
  • No built-in safety filtering compared to API-based alternatives

Key Features

•27 billion parameter architecture
•8,192 token context window
•Text-only chat completion
•Open-source model weights
•Fine-tuning capability
•Local deployment support
•February 2024 knowledge cutoff
•Consumer hardware compatibility

About Gemma 2 27B

Gemma 2 27B is Google's largest model in the open-source Gemma family, positioned as a lightweight option for developers seeking capable performance without the computational requirements of frontier models. With 27 billion parameters, it represents the top tier within Google's Gemma lineup while maintaining accessibility for local deployment and fine-tuning. The model supports text-only interactions through chat completion with an 8,192 token context window. Its February 2024 knowledge cutoff provides relatively current information for training data. As an open-source release, Gemma 2 27B allows developers to download model weights, modify the architecture, and deploy on their own infrastructure without API dependencies. Gemma 2 27B targets use cases where organizations need reasonable language understanding and generation capabilities while maintaining control over their deployment environment. Compared to Google's proprietary Gemini models, it trades some advanced capabilities for transparency and deployment flexibility, making it suitable for applications where open-source licensing and local hosting are priorities.

Common Use Cases

Gemma 2 27B suits applications where organizations need moderate language capabilities with full control over deployment and data privacy. Its open-source nature makes it ideal for fine-tuning on domain-specific datasets, building custom applications without API costs, and scenarios requiring air-gapped or on-premises deployment. The model works well for content generation, summarization, question answering, and conversational interfaces where the 8K context window is sufficient. Organizations choosing between API convenience and deployment control often select Gemma 2 27B when data sovereignty, customization requirements, or long-term cost predictability outweigh the operational complexity of self-hosting.

Frequently Asked Questions

How much does Gemma 2 27B cost per million tokens?

Gemma 2 27B pricing varies by provider and pricing type (standard vs batch). Check the pricing table above for current rates across all providers offering hosted inference.

What is Gemma 2 27B best used for?

Gemma 2 27B excels in applications requiring local deployment, data privacy, and model customization. It's well-suited for content generation, summarization, conversational interfaces, and scenarios where organizations need to fine-tune on proprietary datasets or maintain full control over their AI infrastructure.

Can I run Gemma 2 27B locally instead of using an API?

Yes, Gemma 2 27B is open-source with downloadable model weights, allowing local deployment without API dependencies. The 27B parameter size requires substantial hardware resources but can run on high-end consumer GPUs or server infrastructure, giving you complete control over inference and data handling.