LLM Reference
Google DeepMind

Google DeepMind

Researched today
Flagship Q/$A
Quality52%
$/M out$0.000

178 models across 30 families · Latest: DiffusionGemma 26B A4B IT (2026-06)

Pioneering artificial intelligence research.

CodingRAGAgentsLong contextVisionClassificationJSON / Tool useHighlight

Google DeepMind's portfolio covers 161 active models across 25 non-obsolete families, with task labels spanning coding, rag, and agents. Open a model detail page to compare provider routes and sourced benchmarks.

Portfolio context: 7 decision-task tags, 161 active tracked models, latest research stamp 2026-06-12.

Use it for

  • Teams evaluating coding, rag, and agents across this lab's releases
  • Readers comparing families before locking a flagship SKU
  • 161 tracked SKUs for migration and pricing follow-ups

Do not use it for

  • Choosing a hosting provider without opening a model page for price ladders

Active models

161

Non-deprecated SKUs linked to this researcher

Active families

25

Non-obsolete families in coverage

Open catalog

58 open

18 OSI source / 40 open weights (0 text-match)

Decision task tags

7

Mapped to the site-wide task taxonomy

Latest dated release

2026-06-10

DiffusionGemma 26B A4B IT

Freshness

2026-06-12

Researched today

fresh

Information

Founded2014
London, United Kingdom

Release cadence

Showing 5 recent dated ships (full timeline below). Latest spotlight: DiffusionGemma 26B A4B IT (2026-06-10).

Where this lab wins

  • Coding: 24 tracked models with SWE-bench / HumanEval-style scores.
  • RAG: 45 tracked models with ruler / needle retrieval benchmarks.
  • Agentic: 42 tracked models with BFCL, tau-bench, and SWE-bench tool-use coverage.
  • Long-context: 62 tracked models with context-token or InfiniteBench-class signal.

Flagship quality / price signal

Anchor SKU: Gemma 4 E4B IT (best sourced coding Q/$ in this portfolio).

Coding task grade A · benchmark livecodebench score 52 · cheapest tracked output from ladder $0.000 per 1M tokens

Google DeepMind is a British AI research organization founded in 2014. Pioneering artificial intelligence research. Google DeepMind ships 30 model families totaling 178 models, with the most recent release DiffusionGemma 26B A4B IT in 2026-06. Notable families include DiffusionGemma, Gemma 4, and Gemini 3. Use it as a stable reference for lab background, release coverage, and follow-up model pages as they are. View official API endpoints, benchmark performance, and coding/agent fit for every Google DeepMind model.

About

Google DeepMind is a world-leading AI research organization that has made groundbreaking contributions to the field of AI. Their work on deep learning, reinforcement learning, and game AI has led to remarkable achievements like AlphaGo, which showcased the potential of AI to surpass human performance in complex domains. DeepMind's research also has significant implications for industries such as healthcare, where their AI systems can assist in medical diagnosis and treatment planning. For software engineers and SaaS executives, staying informed about DeepMind's cutting-edge research is essential for understanding the latest advancements and potential applications of AI.

Featured models

ModelReleasedContextInput price ($/1M)Output price ($/1M)LicenseOpenness
DiffusionGemma 26B A4B IT2026-06-10256k--Apache 2.0Open source
Gemma 4 12B2026-06-03256k--Apache 2.0Open source
Gemma 4 12B IT2026-06-03256k--Apache 2.0Open source

Model families

Recent releases

  1. DiffusionGemma 26B A4B IT- 2026-06-10
  2. Gemma 4 12B- 2026-06-03
  3. Gemma 4 12B IT- 2026-06-03
  4. Nano Banana Pro (Gemini 3 Pro Image)- 2026-05-28
  5. Nano Banana 2 (Gemini 3.1 Flash Image)- 2026-05-28

FAQ

Who founded Google DeepMind and when?

Google DeepMind was founded in 2014 and is associated with London, United Kingdom.

What models has Google DeepMind released?

Google DeepMind ships 178 models across 30 families: DiffusionGemma, Gemma 4, and Gemini 3.

Is Google DeepMind's technology open source?

Some Google DeepMind models are open-weight (DiffusionGemma 26B A4B IT, Gemma 4 12B, and Gemma 4 12B IT); others are proprietary (Nano Banana 2 (Gemini 3.1 Flash Image), Gemini 3.1 Flash-Lite, and Gemini 3.1 Flash-Lite).

Where is Google DeepMind headquartered?

Google DeepMind is headquartered in London, United Kingdom.

What is Google DeepMind known for?

Pioneering artificial intelligence research. Its most prominent tracked family is DiffusionGemma.

How can I access Google DeepMind's models?

Google DeepMind's models are available via Google AI Studio, Arcee AI, Alibaba Cloud PAI-EAS, AWS Bedrock, and Bitdeer AI.

Explore related pages

Last reviewed: 2026-06-12. Data sourced from public lab announcements and provider documentation.