Contamination Map Methodology

Text / LLM / Model

Gemma 4

Organisation
Google DeepMind
Family
Gemma · version 4
Released
2026-04-02 (Google blog 'Gemma 4: Byte for byte, the most capable open models' dated Apr 02, 2026 (E2B, E4B, 26B A4B, 31B); 12B announced Jun 03, 2026)
Training cutoff
2025-01-01
Evidence
“Our pre-training dataset is a largescale, diverse collection of data from a wide range of domains and modalities, including web documents, code, images, and audio (for E2B, E4B and 12B), with a cutoff date of January 2025.”
arxiv.org · paper · Section 2.4 Pre-training, Training data · retrieved 2026-10-15
“which includes web documents, code, images, audio, with a cutoff date of January 2025.”
ai.google.dev · model card · Model Data / Training Dataset · retrieved 2026-10-15
Checkpoints
google/gemma-4-E2B, google/gemma-4-E4B, google/gemma-4-12B, google/gemma-4-26B-A4B, google/gemma-4-31B, google/gemma-4-E2B-it, google/gemma-4-E4B-it, google/gemma-4-12B-it, google/gemma-4-26B-A4B-it, google/gemma-4-31B-it
Weights
Open · Apache 2.0
Links
PaperModel card

Training data disclosure: Vague

Level determined for 4% of benchmark cells.

“Our pre-training dataset is a largescale, diverse collection of data from a wide range of domains and modalities, including web documents, code, images, and audio (for E2B, E4B and 12B), with a cutoff date of January 2025.”
arxiv.org · paper · Section 2.4 Pre-training, Training data · retrieved 2026-10-15

Other declared training data

Benchmark cells

Classic LLM benchmarks

Unknown MMLU (Measuring Massive Multitask Language Understanding) · training data not disclosed in enough detail

Unknown GSM8K (Grade School Math 8K) · training data not disclosed in enough detail

Unknown MATH (Mathematics Aptitude Test of Heuristics) · training data not disclosed in enough detail

Unknown HumanEval · training data not disclosed in enough detail

Unknown ARC-Challenge (AI2 Reasoning Challenge, Challenge Set) · training data not disclosed in enough detail

Unknown HellaSwag · training data not disclosed in enough detail

Unknown TruthfulQA · training data not disclosed in enough detail

Unknown WinoGrande · training data not disclosed in enough detail

Multimodal LLM benchmarks

Unknown MMMU (Massive Multi-discipline Multimodal Understanding and Reasoning) · training data not disclosed in enough detail

Unknown MMMU-Pro · training data not disclosed in enough detail

Unknown MathVista · training data not disclosed in enough detail

Unknown ChartQA · training data not disclosed in enough detail

Unknown DocVQA · training data not disclosed in enough detail

Unknown AI2D (AI2 Diagrams) · training data not disclosed in enough detail

Unknown CharXiv · training data not disclosed in enough detail

Recent LLM benchmarks

Unknown GPQA Diamond · training data not disclosed in enough detail

Unknown MMLU-Pro · training data not disclosed in enough detail

Unknown LiveCodeBench (code generation, release_v6) · training data not disclosed in enough detail

Unknown SWE-bench Verified · training data not disclosed in enough detail

Unknown AIME 2024 · training data not disclosed in enough detail

Not seen AIME 2025 · its test items were published from 2025-02-06, after the model's stated training-data cutoff (2025-01-01)

Evidence (4)
“Our pre-training dataset is a largescale, diverse collection of data from a wide range of domains and modalities, including web documents, code, images, and audio (for E2B, E4B and 12B), with a cutoff date of January 2025.”
arxiv.org · paper · Section 2.4 Pre-training, Training data · retrieved 2026-10-15
“which includes web documents, code, images, audio, with a cutoff date of January 2025.”
ai.google.dev · model card · Model Data / Training Dataset · retrieved 2026-10-15
“2025 AIME I problems and solutions. The test was held on Thursday, February 6, 2025.”
web.archive.org · website · retrieved 2026-10-15
“This year the AIME I will take place on February 6, 2025, while the alternate AIME II will take place on February 12, 2025.”
maa.org · website · retrieved 2026-10-15

Unknown MATH-500 · training data not disclosed in enough detail

Notes

Post-training follows Gemma 3's approach plus a thinking mode (report Section 3); post-training data not described beyond filtering. The 12B (released June 2026) shares the model card and January 2025 cutoff. Decontamination statements not shown as cells (they name no benchmark tracked here, or none at all): benchmarks (not named). Accepts image input: “Gemma 4 models are multimodal, handling text and image input (with audio supported on E2B, E4B, and 12B models) and generating text output.”

Sources