Model Details
Gemma 2 27B
Quick answer: Gemma 2 27B is Google's July 2024 open-weight 27B model, scoring 88.6% on ARC-E, 74.0% on GSM8K, and 86.4% on HellaSwag. It was the largest Gemma 2 release and the predecessor to Gemma 3, which adds multimodal capability and a 128K context window.
At a Glance
Where Gemma 2 27B leads
- 86.4% HellaSwag — strong common-sense reasoning at 27B scale
- 88.6% ARC-E — solid scientific knowledge
- Open weights on Hugging Face under Gemma Terms of Use
- Text-focused architecture with efficient inference
- Strong community adoption and fine-tuning support
Where it lags
- 8K context window — very short vs Llama 3.1 (128K) and Gemma 3 27B (128K)
- Superseded by Gemma 3 27B: larger context, multimodal, better benchmarks
- Gemma Terms of Use (not Apache 2.0)
- Text-only; no vision
Best for: Existing Gemma 2 27B integrations; text tasks where short context is acceptable; legacy support.
What Gemma 2 27B Is
Gemma 2 27B (released July 31, 2024) was Google's largest open-weight model in the Gemma 2 family, designed for efficient text generation at 27B scale. It introduced improvements in the attention mechanism and training process compared to Gemma 1.
The model's primary limitation for new deployments is its 8K context window — significantly below Llama 3.1 70B (128K) and Gemma 3 27B (128K). For teams evaluating open-weight 27B models in 2025, Gemma 3 27B is the strongly recommended choice: better benchmarks, 128K context, and multimodal capability.
Gemma 2 27B remains available for teams with existing integrations or who have specifically tuned Gemma 2 models for their use cases.
Specifications
| Field | Value |
|---|---|
| Organization | |
| Parameters | 27B (dense) |
| Context window | 8,192 tokens |
| License | Gemma Terms of Use |
| HuggingFace | google/gemma-2-27b-it |
| Release date | July 31, 2024 |
| Knowledge cutoff | June 2024 |
| Modality | Text only |
Pricing
Gemma 2 27B is available as open weights under the Gemma Terms of Use on Hugging Face — free to self-host.
Public Benchmark Scores
| Benchmark | Score | Source | Date |
|---|---|---|---|
| ARC-E | 88.6% | Benchgen evaluation | 2025-07 |
| GSM8K | 74.0% | Benchgen evaluation | 2025-07 |
| HellaSwag | 86.4% | Benchgen evaluation | 2025-07 |
Gemma 2 27B vs Alternatives
| Model | Context | HellaSwag | Vision | License |
|---|---|---|---|---|
| Gemma 2 27B | 8K | 86.4% | No | Gemma ToU |
| Gemma 3 27B | 128K | — | Yes | Gemma ToU |
| Llama 3.1 70B Instruct | 128K | — | No | Llama 3.1 |
Gemma 3 27B strictly dominates Gemma 2 27B for new deployments: 128K context, multimodal, better benchmarks, same license tier. Migrate to Gemma 3 27B.
Related Models & Benchmarks
- Compare: Gemma 3 27B · Gemma 2 9B · Llama 3.1 70B Instruct
- Benchmarks: HellaSwag · GSM8K
Frequently Asked Questions
What is Gemma 2 27B?
Gemma 2 27B is Google's July 2024 open-weight text model scoring 88.6% ARC-E, 74.0% GSM8K, and 86.4% HellaSwag. It has been succeeded by Gemma 3 27B which adds 128K context and multimodal capability.Should I use Gemma 2 27B or Gemma 3 27B?
Gemma 3 27B for new deployments: 128K context (vs 8K), multimodal vision, and better benchmarks at the same license tier.Specs from Google's official Gemma 2 release (July 2024) and Benchgen evaluations. Last updated 2026-07-24.
Benchmark Leaderboards
This model isn’t on any benchmark leaderboard yet.