Skip to content

Type: AI model

Gemma 4 is Google’s open-weight model family released April 2026 under Apache 2.0, built from the same research as Gemini 3, with context windows up to 256K and native vision and audio. It ships in sizes from compact effective-2B variants up to a dense 31B and a 26B mixture-of-experts. GROUNDING tracks Gemma 4 releases, benchmarks, and on-device-to-server deployment.

Recent Updates

  • 2026-08-12: The Multilingual Quantization Tax: Structural Collapse and Typological Fragility in Edge SLMs (cs.CL updates on arXiv.org) · arxiv.orgOpen Source LLMs Qwen-3.5
  • 2026-08-12: llama.cpp: Running Open-Source Models Locally with Integrated Coding Agent Support (Hacker News) · llama.appOpen Source LLMs Code Agents Alibaba · Google · OpenAI · Hugging Face · Qwen 3.6 · GPT-OSS · Gemma 3
  • 2026-08-13: Practical evaluation of local models: comparing Qwen 3.5/3.6 and Gemma-4 on real coding tasks (Все статьи подряд / Искусственный интеллект / Хабр) · habr.comOpen Source LLMs LLM Evals HuggingFace · OpenAI · Qwen-3.5 · Qwen 3.6 · Claude · Fable · DeepSeek
  • 2026-08-21: mistral.rs: Agentic local LLM inference with Anthropic Messages API support (breakingnewsofficial) · github.comAgents Tool Use Open Source LLMs Anthropic · OpenAI · Hugging Face EricLBuehler · Qwen3-4B Muse Glimmer-30B · DiffusionGemma
  • 2026-08-26: MARS: Multi-Specialist Agent System for Competitive Programming with Retrieval-Augmented Specialist Selection (breakingnewsofficial) · arxiv.orgAgents Code Agents RAG Google
  • 2026-08-31: ReToolSQL: Agentic Reinforcement Learning for Robust Text-to-SQL (breakingnewsofficial) · arxiv.orgAgents Tool Use
  • 2026-08-31: Unsupported Final Claims in Tool-Equipped Language Models: Occurrence and Repair (breakingnewsofficial) · arxiv.orgTool Use Qwen3-32B
  • 2026-09-11: Running OpenCode with Local Ollama Models on Mac (breakingnewsofficial) · tensorsandtokens.comOpen Source LLMs Qwen 3.8

FAQ

What is Gemma 4?

Gemma 4 is Google’s open-weight model family released April 2026 under Apache 2.0, built from the same research as Gemini 3, with context windows up to 256K and native vision and audio. It ships in sizes from compact effective-2B variants up to a dense 31B and a 26B mixture-of-experts. GROUNDING tracks Gemma 4 releases, benchmarks, and on-device-to-server deployment.

What does this page track?

Dated radar mentions, source links, related concepts, and builder-relevant context for Gemma 4, collected automatically by GROUNDING.

When was Gemma 4 last mentioned?

Gemma 4 was most recently mentioned in a radar update dated 2026-09-11.