Gemma 2 27B
by Google · gemma-2 family
27B
parameters
text-generation code-generation reasoning multilingual math creative-writing summarization
Gemma 2 27B is the largest model in Google's Gemma 2 family, delivering top-tier performance among open models in the 20-30B parameter range. It benefits from advanced training techniques including knowledge distillation and RLHF, resulting in highly capable and well-aligned outputs. This model offers excellent performance on reasoning, coding, and creative tasks. While it requires a capable GPU, it fits comfortably on high-end consumer cards at Q4 quantization, making it accessible to enthusiasts with 24GB VRAM GPUs.
Quick Start with Ollama
ollama run 27b-instruct-q4_K_M | Creator | |
| Parameters | 27B |
| Architecture | transformer-decoder |
| Context Length | 8K tokens |
| License | Gemma Terms of Use |
| Released | Jun 27, 2024 |
| Ollama | gemma2:27b |
Quantization Options
| Format | File Size | VRAM Required | Quality | Ollama Tag |
|---|---|---|---|---|
| Q4_K_M recommended | 13.4 GB | 17.7 GB |
★
★
★
★
★
| 27b-instruct-q4_K_M |
| Q5_K_M | 15.7 GB | 20.4 GB |
★
★
★
★
★
| 27b-instruct-q5_K_M |
| Q8_0 | 24.3 GB | 29 GB |
★
★
★
★
★
| 27b-instruct-q8_0 |
Compatible Hardware for Q4_K_M
Showing compatibility for the recommended quantization (Q4_K_M, 17.7 GB VRAM).
Compatible Hardware
5 hardware
device(s) cannot run this model configuration.
Benchmark Scores
75.2
mmlu