reasoning
Gemini 3 Deep Think
Enhanced reasoning mode of Gemini 3 for hard problems in science, research and engineering.
Levelmodel
Classreasoning
Parametersnot in the notes
Contextnot in the notes
Released18 Nov 2025
Opennessproprietary
Runs onnot in the notes
Statusactive
Benchmark Results
vendor-reported
ARC-AGI-2knowledge and reasoning
Updated Deep Think; verified by the ARC Prize Foundation per vendorvendor-reported
84.6
050100
Score in %. Higher is better.
Gemini 3 Deep Think upgrade post
With code execution; ARC Prize Verified per vendorvendor-reported
45.1
050100
Score in %. Higher is better.
Gemini 3 launch post
Codeforces ratingcoding
Updated Deep Thinkvendor-reported
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
GPQA Diamondknowledge and reasoning
vendor-reported
93.8
050100
Score in %. Higher is better.
Gemini 3 launch post
Humanity's Last Exam (HLE)knowledge and reasoning
Without tools; updated Deep Thinkvendor-reported
48.4
050100
Score in %. Higher is better.
Gemini 3 Deep Think upgrade post
Without toolsvendor-reported
41
050100
Score in %. Higher is better.
Gemini 3 launch post
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).
R&D Topics
Identity
- Developer
- Google DeepMind
- Family
- Gemini
- Released
- 18 Nov 2025 sourcesannouncement page
- Weights published
- –
- Record checked
- 11 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
- consumer app, gated preview
Where this comes from
The licence file governs over registry tags and aggregator labels.
Deployment
- Runs on
- not in the notes
- Footprint
- Specialized reasoning mode of Gemini 3; Google AI Ultra subscribers in the Gemini app, early access via Gemini API by expression of interest. Major upgrade on 12 February 2026
- 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.
Lineage
The notes record no lineage for this model.
Industry Use Cases
No industry use case links this model.
Links
ModelGemini 3 Deep ThinkGoogle DeepMind, reasoning
Gemini 3 Deep ThinkGoogle DeepMind, reasoningR&D TopicsReasoning via Large-Scale RL
Use CasesChinese Medical LLMsHealthcareUse cases of models in the same R&D topic; none is recorded for this model itself.
Who Runs Itnone recordedDeployments of these use cases; none names this model.
Organisations 1
Sources
- blog.google/products/gemini/gemini-3 https://blog.google/products/gemini/gemini-3/
- blog.google/innovation-and-ai/models-and-research/gemini-... https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-deep-think/
To cite this page: Project AURION, AI Use-Cases Research & Innovation, "Gemini 3 Deep Think", https://aurion.semai.id/models/google-deepmind-gemini-3-deep-think/, records of October 2026.