Google model product
Gemma 4 12B is Google DeepMind's encoder-free multimodal instruction-tuned model with 11.95 billion parameters and a 256K context window. It supports text, image, audio, and video inputs with text output, projecting image patches and audio waveforms directly into a single decoder-only transformer for streamlined local deployment.
Updated Aug 10, 2026. Default version: Gemma 4 12B
Structured fields from the published default version.
This profile uses the latest version under this unique model that has a calculated LLMBoard score. Arena and price are excluded.
Published benchmark records for the scored version Gemma 4 12B.
| BIG-Bench Extra Hard | 0.5 | 3 | 11 | 80.0% | C | |
| CoVoST2 | 0.4 | 3 | 4 | 33.3% | C | |
| FLEURS | 0.9 | 4 | 6 | 40.0% | C | |
| MedXpertQA | 0.5 | 8 | 12 | 36.4% | C | |
| MRCR v2 (8-needle) | 0.4 | 9 | 21 | 60.0% | C | |
| MathVision | 0.8 | 13 | 32 | 61.3% | C | |
| AIME 2026 | 0.8 | 14 | 17 | 18.8% | C | |
| OmniDocBench 1.5 | 0.2 | 14 | 16 | 13.3% | C | |
| CodeForces | 0.6 | 15 | 16 | 6.7% | C | |
| LiveCodeBench v6 | 0.7 | 26 | 53 | 51.9% | C | |
| MMMLU | 0.8 | 35 | 49 | 29.2% | C | |
| MMMU-Pro | 0.7 | 37 | 65 | 43.8% | C | |
| MMLU-Pro | 0.8 | 65 | 129 | 50.0% | C | |
| Humanity's Last Exam | 0.1 | 88 | 92 | 4.4% | C | |
| GPQA | 0.8 | 92 | 233 | 60.8% | C |
Preference and agent-evaluation signals from published Arena datasets.
The default version has no published Arena rows, or its source alias has not been resolved.
Official vendor API PAYG pricing is summarized first. The table then lists individual provider offerings without treating their minimum as the official price.
| NanoGPT | gemma-4-12b-it | global | $0.06 | $0.30 | 262.1K |
Official prices use only the vendor's configured official Provider and positive standard USD PAYG rates. Third-party offers remain explicitly labeled.
All published versions linked to this unique model. The score columns identify the version used by the current overall ranking.
| Gemma 4 12B | 48.1 | 12B | N/A | N/A | No | Apache 2.0 |
A concise description based on the published model registry.
Gemma 4 12B is Google DeepMind's encoder-free multimodal instruction-tuned model with 11.95 billion parameters and a 256K context window. It supports text, image, audio, and video inputs with text output, projecting image patches and audio waveforms directly into a single decoder-only transformer for streamlined local deployment.
Use the benchmark, Arena and pricing sections above as separate evidence. A missing field means the current data snapshot does not support that claim.
Data snapshot: 2026-08-07. Editorial model content is not available in the backend.
Open a comparison with the three ranked models immediately above and below this model.
Recommendations prioritize the same model type and family, then the closest published LLMBoard score.
Common questions about Gemma 4 12B.
Gemma 4 12B's default version was released on May 23, 2026.
No official standard PAYG price is currently available for Gemma 4 12B. The lowest tracked third-party offer starts at $0.06 input and $0.30 output via NanoGPT.
Gemma 4 12B is published under Google in the model registry.
The current registry does not publish a context window for the default version.
No. The default version is not marked as having publicly available weights.
1 published provider offerings are linked to the default version.
Nearby ranked alternatives include Llama 4 Maverick, MiniMax M1 40K, Qwen3.6 35B A3B.