Google releases Gemma 4 multimodal open-source model family
Original title:Google has released Gemma 4, a new family of multimodal open-weight models including Gemma 4 E2B, Ge...
AI summary
Google DeepMind launched the Gemma 4 family of four multimodal open-source models supporting text, image and video input. The 31B (dense) and 26B A4B (MoE) models have 256k context windows and can run on a single H100; the two smaller models support 128k context. On GPQA Diamond, Gemma 4 31B (Reasoning) scored 85.7%, second only to Qwen3.5 27B, while using only about 1.2M output tokens for better efficiency; 26B A4B (Reasoning) scored 79.2%, surpassing gpt-oss-120B.
Key Takeaways
- Google DeepMind launched the Gemma 4 family of four multimodal open-source models supporting…
- The 31B (dense) and 26B A4B (MoE) models have 256k context windows and can run on a single H100
- the two smaller models support 128k context.
- On GPQA Diamond,Gemma 4 31B (Reasoning) scored 85.
- 7%,second only to Qwen3.
Quick Facts
- Published
- Updated
- Category
- Other
Google发布多模态开源模型Gemma 4,单卡H100可跑且科学推理能力突出
Related entities
Source & copyright
This article is based on curated titles and AI summaries from the AIHOT public API, with attribution and links back. Original copyright remains with the original publishers and authors; this site does not mirror full texts.
How to Cite
When citing this article, keep the title, date, and permalink, and credit Orange News as the source.
Orange News: Google releases Gemma 4 multimodal open-source model family. April 3, 2026. https://new.nat6.net/en/news/cmnw1ypby00ocslc36i8pvvx8/