Type: AI model
Gemma 3 is a generation of Google’s open-weight Gemma family, offered in sizes including a 27B instruct variant and available on Hugging Face. It preceded Gemma 4 and added multimodal and long-context capabilities. GROUNDING tracks Gemma 3’s variants, benchmarks, and open-model adoption.
Recent Updates
- 2026-06-30: LLM Confidence Reports Track Commitment More Than Correctness (cs.LG updates on arXiv.org) · arxiv.org — LLM Evals Dharshan Kumaran Gemma 4
- 2026-06-30: Study finds multilingual fine-tuning can change safety behavior in uneven ways (cs.CL updates on arXiv.org) · arxiv.org — LLM Evals Llama 3.2 · Qwen3
- 2026-07-01: Bangla event detection benchmark tests robustness under noisy text (cs.CL updates on arXiv.org) · arxiv.org — LLM Evals arXiv Tanvir Ahmed Sijan BanglaBERT XLM-R · LLaMA-3
- 2026-07-03: LOCOS Detects Non-Literal Retrieval Heads in Long-Context Models (cs.CL updates on arXiv.org) · arxiv.org — Long Context Qwen3 OLMo-3.1
- 2026-07-07: Gemma 4 Technical Report highlights multimodal, long-context claims (alphaXiv) · twitter.com — Open Source LLMs Long Context LLM Evals Google · alphaXiv · Gemma 4 · E2B E4B
FAQ
What is Gemma 3?
Gemma 3 is a generation of Google’s open-weight Gemma family, offered in sizes including a 27B instruct variant and available on Hugging Face. It preceded Gemma 4 and added multimodal and long-context capabilities. GROUNDING tracks Gemma 3’s variants, benchmarks, and open-model adoption.
What does this page track?
Dated radar mentions, source links, related concepts, and builder-relevant context for Gemma 3, collected automatically by GROUNDING.
When was Gemma 3 last mentioned?
Gemma 3 was most recently mentioned in a radar update dated 2026-07-07.
Category: Text / Language Models