Gemma 2 27B IT (Together)
Gemma 2 27B IT is Google DeepMind's largest instruction-tuned Gemma 2 model, hosted on Together AI. The 27B scale extends the architectural improvements of the Gemma 2 series — interleaved local and global attention, logit soft-capping, and distillation-enhanced training — to deliver performance that competes with significantly larger models on standard benchmarks.
For teams needing stronger open-weight capability than the 9B variant without moving to a 70B model, the 27B sits in a practical sweet spot. Together AI makes it accessible for demanding text tasks including long-document analysis, nuanced summarization, and complex multi-step reasoning, at a cost profile that remains workable for production use.
Key Features
Near-frontier reasoning performance for an open-weight model at this scale
Interleaved local and global attention for efficient long-context handling
High-quality instruction following for complex, multi-step directives
Strong summarization of lengthy technical and legal documents
Reliable structured output generation (JSON, tables, outlines)
Commercial-permissive license enabling fine-tuning and self-hosting
Ideal Use Cases
Long-document analysis and executive summary generation
Complex multi-turn assistants requiring strong contextual memory
Structured data extraction from unstructured text at scale
Open-weight alternative to proprietary APIs for regulated industries
Research and academic NLP applications needing reproducible baselines
Example Prompts for Gemma 2 27B IT (Together)
Technical Specifications
| Provider | Together |
| Category | Text |
| Modality | Text -> Text |
| Context Window | 8192 tokens |
Frequently Asked Questions
Try Gemma 2 27B IT (Together) now
Start using Gemma 2 27B IT (Together) instantly — 100 free credits, no credit card required. Access 750+ AI models through one platform.