Gemma 2 27B
Language modelActive
Gemma 2 27B
- Released
- June 27, 2024
- Data date
- September 8, 2026
- Successor
- Gemma 3 27B
- Price
- Input · Output / 1M tokens
- Input
- Text
- Output
- Text
Gemma 2 27B is the largest model in the Gemma 2 family. Google trained the dense text model on 13 trillion tokens, compared with 8 trillion for 9B and 2 trillion for 2B. Training covered primarily English web text, code, and mathematics.
The weights use the Gemma Terms of Use. Gemma 3 27B later adds native image input; Gemma 2 processes text only.
Specifications and access
| Specification | Value and source |
|---|---|
| Weights | Open weightsSource |
| Access | API, LocalSource |
| Parameters | 27BSource |
| Context window | 8,192 tokensSource |
| Release | 2024-06-27Source |
| Parameters | 27 BSource |
| Input | TextSource |
| Output | TextSource |
| Architecture | Decoder-only transformer; English text-to-textSource |
| License | Gemma Terms of UseSource |
| Training tokens | 13,000,000,000,000 tokensSource |
| Languages | EnglishSource |
Published benchmarks
LegalBench
Overall
higher is betterAs of September 29, 2026Scale: 0% - 100%
Methodology and context
- Comparison cohort
- vals-legal-bench:1:overall
- Metric
- accuracy
- Scale
- 0-100
- Participants
- 78
- Methodology and context
- Source model ID: together/google/gemma-2-27b-it. Source model name: Gemma 2 27B. Evaluator: Vals AI. Dataset type: public. Leaderboard date: 2026-09-27. Latency in seconds: 0.887. Source ID: vals-ai. Evidence capture ID: vals-legal_bench-2026-09-29-c1213019b552. Source payload hash: c1213019b55234a262ae86fd4ebfe291530f315c988a4ee32fd501c13777f85f. Parser version: vals-astro-v2. Imported at: 2026-09-29
Measurements without matching peer values
These measurements have no matching peer values under the same test conditions. Their original values and sources remain available here.
MMLU (Show measurement, test conditions, and source)
- Source value
- 76.2
- Score
- 76.2
- Metric
- percent
- Unit
- %
- Category
- vendor-evaluation
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Google DeepMind
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Gemma 2 technical report Table 17; instruction-tuned 27B; few-shot benchmark
MBPP (Show measurement, test conditions, and source)
- Source value
- 67.4
- Score
- 67.4
- Metric
- percent
- Unit
- %
- Category
- vendor-evaluation
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Google DeepMind
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Gemma 2 technical report Table 17; instruction-tuned 27B; few-shot benchmark
Recorded measurement details
LegalBench: Overall (Show measurement, test conditions, and source)
- Source value
- 67.978
- Score
- 67.978
- Metric
- accuracy
- Unit
- %
- Benchmark version
- 1
- Category
- legal
- Task
- overall
- Task label
- Overall
- Direction
- Higher is better
- Comparison cohort
- vals-legal-bench:1:overall
- Source type
- independent-evaluator
- Evaluator
- Vals AI
- Status
- active
- Retrieved at
- 2026-09-29
- Methodology
- Source model ID: together/google/gemma-2-27b-it. Source model name: Gemma 2 27B. Evaluator: Vals AI. Dataset type: public. Leaderboard date: 2026-09-27. Latency in seconds: 0.887. Source ID: vals-ai. Evidence capture ID: vals-legal_bench-2026-09-29-c1213019b552. Source payload hash: c1213019b55234a262ae86fd4ebfe291530f315c988a4ee32fd501c13777f85f. Parser version: vals-astro-v2. Imported at: 2026-09-29
Our tests
SVG short test
One SVG prompt and fixed output settings. Captures are shown only with 0 reported reasoning tokens or documented lack of reasoning controls. Identical test conditions do not make the models' internal compute identical.
Popularity
31.3 / 100
Predecessor and successor
Sources and data date
| Type | Evidence and data date |
|---|---|
| Price | Base API prices without caching or batch discounts. Higher context tiers and other rates are listed under Costs. |
| Input | Documented modalities. Retrieved September 8, 2026. |
| Output | Documented modalities. Retrieved September 8, 2026. |
| Research status | Research date September 8, 2026. 4 source URLs checked. This documents the inspected sources, not an exhaustive inventory of every publication. |
| Not found in the inspected sources | active-parameters-billions Original wording Active parameter count is not separately published for this dense model. |
| Not found in the inspected sources | api-model-id Original wording Gemma is distributed as open weights; the card does not define a hosted Gemini API model ID. |
| Not found in the inspected sources | context-tokens Original wording The Gemma 2 model card does not state a context-window size. |
| Not found in the inspected sources | knowledge-cutoff Original wording The Gemma 2 model card does not state a training-data knowledge cutoff. |
| Unresolved | AA benchmark rows Original wording No exact Artificial Analysis model page located in the checked public sitemap. |
| Additional source | Model metadata · Gemma 2 27B instruction-tuned weights · Editorial description reviewed October 2, 2026 |
| Additional source | Context window data |
| Additional source | Vals AI (retrieved September 29, 2026) |
| Additional source | Google DeepMind (retrieved September 8, 2026) |
| Additional source | Model metadata (retrieved September 8, 2026) |
| Additional source | OpenRouter: Popularity index (derived), CC BY 4.0 |
| Additional source | Vercel AI Gateway: Popularity index (derived), CC BY 4.0 |
| Additional source | Creative Commons Attribution 4.0 |
| Additional source | Google Gemma 3 model card (2026-09-19) |