multimodal LLMfamily
Gemini
Google DeepMind's proprietary model family, from Gemini 2.5 through Gemini 3.x Flash to Gemini 4 Argon.
Levelfamily
Classmultimodal LLM
Parametersnot in the notes
Contextnot in the notes
Releasednot in the notes
Opennessproprietary
Runs onnot in the notes
Statusactive
Benchmark Results
vendor-reportedmaintainer-runindependent
ARC-AGI-2knowledge and reasoning
Gemini 3 Deep ThinkUpdated Deep Think; verified by the ARC Prize Foundation per vendorvendor-reported
84.6
Gemini 3 Deep Think upgrade post
45.1
050100
Score in %. Higher is better.
Gemini 3 launch post
Arena Text leaderboardhuman preference arena
1,487 ±3
1,2001,4001,600
Score in Elo; the axis starts at 1,200, as Elo has no zero. Higher is better.
rank 17, Arena Text overall
Arena Vision leaderboardhuman preference arena
1,297 ±10
rank 6 of 165, Arena Vision overall, style control default, 165 models
1,289 ±11
1,2701,2901,310
Score in Elo; the axis starts at 1,270, as Elo has no zero. Higher is better.
rank 14 of 165, Arena Vision overall, style control default, 165 models
Artificial Analysis Intelligence Index v4.3.2composite index
57
rank 1, DataLearner mirror of AA Index v4.0
53
rank 8 of 205, AA leaderboard, 205 current models
41
04590
Score in index. Higher is better.
AA leaderboard, 205 current models
AutomationBenchagents and tool use
benchlm.ai frontier compositecomposite index
81.8
04590
Score in index. Higher is better.
benchlm.ai frontier top 6
Codeforces ratingcoding
3,455
3,4503,4553,460
Score in Elo; the axis starts at 3,450, as Elo has no zero. Higher is better.
Gemini 3 Deep Think upgrade post
DeepSWE v1.1coding
GPQA Diamondknowledge and reasoning
Humanity's Last Exam (HLE)knowledge and reasoning
48.4
Gemini 3 Deep Think upgrade post
41
050100
Score in %. Higher is better.
Gemini 3 launch post
LABBench 2 2domain
MMMUmultimodal
Gemini 3 Provendor-reported
87.2
050100
Score in %. Higher is better.
rank 1, BenchLM aggregator, top score
OmniDocBench v1.6multimodal
92.91
050100
Score in score. Higher is better.
PaddleOCR-VL-1.6 paper table
Real5-OmniDocBenchmultimodal
89.24
050100
Score in score. Higher is better.
PaddleOCR-VL-1.6 paper table
SWE-bench Procoding
46.1 ±3.60
050100
Score in %. Higher is better.
SWE-bench Verifiedcoding
Vals Indexcomposite index
Where this comes from
Every score names who ran it: vendor-reported, maintainer-run or independent. A rank is shown with the population it was ranked in (gap rule 3).
Lineage
product ofsuccessor
16 members have no release date in the notes, so they are not on this timeline.
AMIE is a child (product of)high confidencepress claimGemini 4 Argon belongs to this familyLLM30 Sep 2026Gemini 3.8 Flash belongs to this familyLLM2 Sep 2026Gemini 3.7 Flash belongs to this familyLLMAug 2026Gemini 3.6 Flash belongs to this familyLLM21 Jul 2026Gemini 3.5 Flash belongs to this familymultimodal LLMMay 2026Gemini 3.1 Pro (preview) belongs to this familymultimodal LLM19 Feb 2026Gemini 3 Deep Think belongs to this familyreasoning18 Nov 2025Gemini 3 Pro belongs to this familymultimodal LLM2025antigravity-preview-09-2026 belongs to this familyLLMdeep-research-max-preview-04-2026 belongs to this familyLLMdeep-research-preview-04-2026 belongs to this familyLLMGemini 2.5 Flash belongs to this familymultimodal LLMGemini 2.5 Flash-Lite belongs to this familyLLMGemini 2.5 Pro belongs to this familyLLMGemini 3.1 Flash-Lite belongs to this familyLLMGemini 3.5 Flash-Lite belongs to this familymultimodal LLMGemini 3.5 Pro belongs to this familyLLMGemini 3.8 Flash Cyber belongs to this familyLLMGemini Nano belongs to this familyedge LLMgemini-2.5-computer-use-preview-10-2025 belongs to this familyLLMgemini-3.8-live belongs to this familyspeech synthesisgemini-3.8-live-extended-thinking belongs to this familyspeech synthesisgemini-robotics-er-2-preview belongs to this familymultimodal LLMWhere this comes from
Parent and child relations from model cards, Hugging Face base_model fields and the genealogy tables of the research notes. The tree shows two generations each way, 3 models. Confidence and evidence are the notes' own.
Industry Use Cases
R&D Topics
Deployments
XLSmart (XL, AXIS, Smartfren)productionIndonesia21 Jul 2026
XLSmart targets 20 to 30 percent Gemini adoption
Consumer AI assistant bundlingcompany statementgizmologi.id
Indosat Ooredoo Hutchison (IM3, Tri)productionIndonesia8 Apr 2026
Limited from February 2026, expanding
Consumer AI assistant bundlingcompany statementkoran-jakarta.com
HoneywellplannedUSA5 Nov 2024
First solutions due 2025; 2026 status unverified
Industrial copilots and maintenance assistantscompany statementdesignnews.com
PT Pupuk Indonesia with DevoteamproductionIndonesia
Extraction time from 5 to 10 minutes down to 40 to 70 seconds (claim)
Operational document processing and data entrycompany statementcloud.google.comWhere this comes from
From G13, the deployment rows of T08 to T10, and the Indonesian-language search of 11 October 2026. An outcome claim is the organisation's own unless the evidence type says independent report.
Identity
- Developer
- Google DeepMind, Google
- Family
- not in the notes
- Released
- – sources
- Weights published
- –
- Record checked
- 10 Oct 2026
Where this comes from
Developer and dates from the research notes. Release date is the vendor's public announcement; weights date is when open weights were published.
Architecture and Size
- Architecture
- not in the notes
- Total parameters
- – sources
- Active parameters
- –
- Checkpoint parameters
- –
- Context length
- not in the notes
- Input
- not in the notes
- Output
- not in the notes
Where this comes from
Parameter counts as the developer states them. Where the checkpoint holds a different total, both are shown (register A16).
Openness and Access
- Openness
- proprietary
- Licence
- Proprietary sources
- Access routes
- not in the notes
Where this comes from
The licence file governs over registry tags and aggregator labels.
Deployment
- Runs on
- not in the notes
- Footprint
- not in the notes
- Runtimes
- not in the notes
Where this comes from
Deployment tiers and runtimes as named in the notes. Footprint is the notes' figure, not a measurement by AURION.
Indonesia
- Indonesian support
- not stated
- Regional languages
- not in the notes
- Jakarta hosting
- unknown
- Languages
- not in the notes
Where this comes from
Indonesia fields are carried on every model; unknown is an allowed value.
Links
ModelGeminiGoogle DeepMind, multimodal LLM
GeminiGoogle DeepMind, multimodal LLMR&D TopicsFrontier Pretraining Scaling
Models 24
AMIEGoogle, Google DeepMind, multimodal LLMProduct ofGemini 3.6 FlashGoogle DeepMind, LLMFamilyGemini 3 ProGoogle DeepMind, Google, multimodal LLMFamilyGemini 3.8 FlashGoogle DeepMind, LLMFamilyGemini 2.5 FlashGoogle DeepMind, Google, multimodal LLMFamilyGemini 3.1 Pro (preview)Google DeepMind, Google, multimodal LLMFamilyGemini 3.5 FlashGoogle DeepMind, multimodal LLMFamilyGemini 3.7 FlashGoogle DeepMind, LLMFamilyGemini NanoGoogle, edge LLMFamilygemini-2.5-computer-use-preview-10-2025Google DeepMind, LLMFamilyGemini 2.5 ProGoogle DeepMind, LLMFamilyGemini 3.8 Flash CyberGoogle DeepMind, LLMFamily+12
Gemini 4 ArgonGoogle DeepMind, LLMFamilyantigravity-preview-09-2026Google DeepMind, LLMFamilydeep-research-max-preview-04-2026Google DeepMind, LLMFamilydeep-research-preview-04-2026Google DeepMind, LLMFamilyGemini 2.5 Flash-LiteGoogle DeepMind, LLMFamilyGemini 3 Deep ThinkGoogle DeepMind, reasoningFamilyGemini 3.1 Flash-LiteGoogle DeepMind, LLMFamilyGemini 3.5 Flash-LiteGoogle DeepMind, multimodal LLMFamilyGemini 3.5 ProGoogle DeepMind, LLMFamilygemini-3.8-liveGoogle DeepMind, speech synthesisFamilygemini-3.8-live-extended-thinkingGoogle DeepMind, speech synthesisFamilygemini-robotics-er-2-previewGoogle DeepMind, multimodal LLMFamilyDeployments 4
HoneywellGemini on Vertex AI; Gemini Nano at the edge planned, USAIndosat Ooredoo Hutchison (IM3, Tri)Google Gemini AI Plus bundled with IM3 and Tri data packages, IndonesiaPT Pupuk Indonesia with DevoteamGoogle Vision AI and Gemini for document data extraction, IndonesiaXLSmart (XL, AXIS, Smartfren)Paket Plus Google Gemini: six months of Gemini with data packages, Indonesia
Sources
- Models | Gemini API | Google AI for Developers ai.google.dev https://ai.google.dev/gemini-api/docs/models
- AMIE, our research medical AI system, demonstrates real-time clinical video consultation capabilities in a first-of-its-kind study. blog.google https://blog.google/innovation-and-ai/models-and-research/google-research/amie-video-consultations/
- designnews.com/automation/honeywell-uses-google-cloud-ai-to-... https://www.designnews.com/automation/honeywell-uses-google-cloud-ai-to-accelerate-autonomous-operations
- ai.google.dev/gemini-api/docs/models/gemini-3.1-pro-preview https://ai.google.dev/gemini-api/docs/models/gemini-3.1-pro-preview
- ai.google.dev/gemini-api/docs/models/gemini-2.5-flash https://ai.google.dev/gemini-api/docs/models/gemini-2.5-flash
To cite this page: Project AURION, AI Use-Cases Research & Innovation, "Gemini", https://aurion.semai.id/models/google-deepmind-gemini/, records of October 2026.