Gemma 4 (31B dense)
Released 2026-04 · chat · 256K tokens · 4 benchmarks · Open weight
Editorial notes
Lanzado 2 abril 2026. Variante flagship 31B dense de la familia Gemma 4 (Apache 2.0). 256K context, 140+ idiomas. 'Byte for byte, los modelos open mas capaces' segun Google. Compite con modelos 20x mas grandes.
Spec sheet
- Empresa
- Google DeepMind
- Pais
- US
- Tipo
- chat
- Release
- 2026-04
- Context
- 256K tokens
- Licencia
- apache-2.0
- Quants
- FP16, Q8_0, Q5_K_M, Q4_K_M, Q3_K_M
- Slug
- gemma-4
Quick install
3 toolsollama run gemma4:27bNota: ~16 GB VRAM en Q4_K_M
Buscar "gemma-4-26B-A4B" en LM Studio model browservllm serve google/gemma-4-26b-a4b-it --tensor-parallel-size 1Nota: Requiere ~52 GB VRAM en FP16
Benchmarks (4)
Reasoning 2
Cite this model
BibTeX · APA
BibTeX
@misc{frontier-gemma-4,
title = {Gemma 4 (31B dense)},
author = {{Google DeepMind}},
year = {2026},
note = {Frontier Benchmarks AI atlas. Accessed 2026-05-08},
url = {https://frontierbenchmarks.com/models/gemma-4}
} APA
Google DeepMind (2026). Gemma 4 (31B dense) [Large language model]. Frontier Benchmarks AI. Retrieved 2026-05-08, from https://frontierbenchmarks.com/models/gemma-4
Citation refleja la pagina del atlas, no el paper original del modelo. Para el paper, ve a la seccion "Recursos" arriba.