Benchmark Almanac

Google DeepMind

Best on the Capabilities Index Gemini 4 Argon 166.1
76 models 44 model cards read deepmind.google

Google DeepMind's models against the field

Every model on the Epoch Capabilities Index by release date, with Google DeepMind's in color.

50 75 100 125 150 175 2023 2024 2025 2026 2027 Gemini 4 Argon Gemini 3.8 Flash Gemini 3 Flash 50 75 100 125 150 175 2023 2024 2025 2026 2027 Gemini 4 Argon Gemini 3.8 Flash

Its strongest models

Google DeepMind Sep 2026

Gemini 4 Argon

Closed weights Flagship

166.1 #7 of 271
Epoch Capabilities Index
  • AutomationBench 51.3% #2 /5
  • Agents' Last Exam 40.2% #6 /12
  • MultiChallenge 78.7% #2 /35
Google DeepMind Aug 2026

Gemini 3.7 Flash

Closed weights Mid-size

158.2 #21 of 271
Epoch Capabilities Index
  • DeepSWE 65.5% #12 /28
  • Terminal-Bench 4.0 11.2% #23 /23
  • Harvey LAB (Legal Agent Benchmark) 8.75% #25 /80
Google DeepMind Sep 2026

Gemini 3.8 Flash

Closed weights Mid-size

157.6 #23 of 271
Epoch Capabilities Index
  • Humanity's Last Exam 44.5% #5 /47
  • DeepSWE 73.8% #2 /28
  • Terminal-Bench 4.0 19.1% #17 /23
Google DeepMind Feb 2026

Gemini 3.1 Pro

Closed weights Flagship

155.2 #37 of 271
Epoch Capabilities Index
  • Humanity's Last Exam 46.4% #3 /47
  • SWE-bench Verified 75.6% #12 /32
  • SWE-Bench Pro 46.1% #5 /24
Google DeepMind May 2026

Gemini 3.5 Flash

Closed weights Mid-size

155.0 #40 of 271
Epoch Capabilities Index
  • SWE-bench Verified 79.3% #3 /32
  • MMMU-Pro 83.6% #2 /72
  • DeepSWE 36.1% #25 /28
Google DeepMind Jul 2026

Gemini 3.6 Flash

Closed weights Mid-size

154.7 #42 of 271
Epoch Capabilities Index
  • DeepSWE 46.7% #23 /28
  • ZeroBench 18% #17 /66
  • Harvey LAB (Legal Agent Benchmark) 3.33% #43 /80

What its model cards measure

Tall ticks are Google DeepMind model cards that list the benchmark in their headline table. Gray dots are cards that leave it out. Rows are the benchmarks the lab listed most, at least twice each, in the order it started using them.

MATH HellaSwag BIG-Bench Hard DROP GSM8K WinoGrande PIQA ARC-Challenge MMMU MMLU HumanEval GPQA Diamond EgoSchema HiddenMath FACTS Grounding MMLU-Pro Vibe-Eval MRCR (Michelangelo, original) Global MMLU Lite LiveCodeBench v5 SimpleQA Humanity's Last Exam AIME 2025 SWE-bench Verified MRCR v2 MMMLU MMMU-Pro CharXiv

Model cards

DateModelDocumentBenchmarks
28 Oct 2026 Gemini 4 Carbon Gemini 4 Carbon: our next frontier model (Google blog) 19
13 Oct 2026 Gemini 4 Argon Gemini 4 Argon model card 18
15 Sep 2026 Gemini 3.8 Live Extended Thinking Introducing Gemini 3.8 Live and 3.8 Live Extended Thinking (Google blog) 3
2 Sep 2026 Gemini 3.8 Flash Cyber Introducing Gemini 3.8 Flash and 3.8 Flash Cyber (Google blog) 3
2 Sep 2026 Gemini 3.8 Flash Gemini 3.8 Flash model card 13
13 Aug 2026 Gemini 3.7 Flash Gemini 3.7 Flash model card 18
30 Jul 2026 Gemini Robotics ER 2 Introducing Gemini Robotics ER 2 (Google blog) 7
21 Jul 2026 Gemini 3.5 Flash Cyber Introducing Gemini 3.5 Flash Cyber (Google DeepMind blog) 3
21 Jul 2026 Gemini 3.5 Flash-Lite Gemini 3.5 Flash-Lite model card 7
21 Jul 2026 Gemini 3.6 Flash Gemini 3.6 Flash model card 8
10 Jun 2026 DiffusionGemma 26B A4B DiffusionGemma model card (Google AI for Developers) 14
19 May 2026 Gemini 3.5 Flash Gemini 3.5 Flash model card 13
14 Apr 2026 Gemini Robotics-ER 1.6 Gemini Robotics ER 1.6: Enhanced Embodied Reasoning (Google DeepMind blog) 3
2 Apr 2026 Gemma 4 26B A4B Gemma 4 model card (Google AI for Developers) 14
2 Apr 2026 Gemma 4 31B IT Gemma 4 model card 16
3 Mar 2026 Gemini 3.1 Flash-Lite Gemini 3.1 Flash-Lite model card 10
19 Feb 2026 Gemini 3.1 Pro Gemini 3.1 Pro model card 16
12 Feb 2026 Gemini 3 Deep Think Gemini 3 Deep Think: Advancing science, research and engineering (Google blog) 7
17 Dec 2025 Gemini 3 Flash Gemini 3 Flash model card 21
18 Nov 2025 Gemini 3 Pro Gemini 3 Pro model card 20
7 Oct 2025 Gemini 2.5 Computer Use Introducing the Gemini 2.5 Computer Use model (Google blog) 3
25 Sep 2025 Gemini 2.5 Flash-Lite Sep 2025 Gemini 2.5 Flash-Lite Model Card, Preview (09-2025) columns (Google DeepMind) 11
25 Sep 2025 Gemini 2.5 Flash Sep 2025 Continuing to bring you our latest models, with an improved Gemini 2.5 Flash and Flash-Lite release (Google Developers Blog) 13
14 Aug 2025 Gemma 3 270M Introducing Gemma 3 270M (Google Developers Blog) and Gemma 3 model card 6
1 Aug 2025 Gemini 2.5 Deep Think Gemini 2.5 Deep Think model card 4
26 Jun 2025 Gemma 3n E4B Introducing Gemma 3n: The developer guide (Google Developers Blog) and Gemma 3n model card 28
17 Jun 2025 Gemini 2.5 Flash Jun 2025 Gemini 2.5 Flash Model Card, GA Thinking column (Google DeepMind) 12
17 Jun 2025 Gemini 2.5 Flash-Lite Jun 2025 Gemini 2.5 Flash-Lite Model Card (Google DeepMind) 12
17 Jun 2025 Gemini 2.5 Pro Jun 2025 Gemini 2.5 Pro model card (updated 2025-06-27) 14
6 May 2025 Gemini 2.5 Pro May 2025 Gemini 2.5 Pro Preview: even better coding performance (Google Developers Blog) 2
17 Apr 2025 Gemini 2.5 Flash Apr 2025 Start building with Gemini 2.5 Flash (Google Developers Blog) 11
25 Mar 2025 Gemini 2.5 Pro Mar 2025 Gemini 2.5: Our most intelligent AI model (blog.google launch post) 12
12 Mar 2025 Gemma 3 27B Gemma 3 model card (ai.google.dev, last updated 2025-08-14) 50
5 Feb 2025 Gemini 2.0 Flash-Lite Gemini 2.0 is now available to everyone (Google blog) 13
5 Feb 2025 Gemini 2.0 Flash Feb 2025 Gemini 2.0 is now available to everyone (Google blog) 13
5 Feb 2025 Gemini 2.0 Pro Gemini 2.0 is now available to everyone (Google blog) 13
21 Jan 2025 Gemini 2.0 Flash Thinking Jan 2025 Gemini 2.0 Flash Thinking (Google DeepMind model page) 3
11 Dec 2024 Gemini 2.0 Flash Dec 2024 Introducing Gemini 2.0 launch post (blog.google) 13
27 Jun 2024 Gemma 2 27B Gemma 2 model card (ai.google.dev) 17
14 May 2024 Gemini 1.5 Pro May 2024 Gemini 1.5 technical report, May 2024 update with 1.5 Flash (arXiv 2403.05530v3) 36
21 Feb 2024 Gemma 7B Gemma model card (ai.google.dev) 18
15 Feb 2024 Gemini 1.5 Pro Feb 2024 Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context (technical report, Feb 2024 version = arXiv 2403.05530v1) 26
6 Dec 2023 Gemini 1.0 Ultra Gemini: A Family of Highly Capable Multimodal Models (technical report, arXiv 2312.11805v1) 17
10 May 2023 PaLM 2-L PaLM 2 Technical Report (arXiv 2305.10403v1) 32

Every Google DeepMind model