Mathematical problem solving
Explore proofs, counterexamples, derivations, and alternative solution paths when a normal chat response is too shallow.
Independent tool overview
Gemini 3.1 Deep Think at a glance
Gemini 3.1 Deep Think is Google's highest reasoning level in the Gemini app, using more parallel inference-time reasoning for difficult math, science, engineering, and coding problems.
Visit the official Gemini 3.1 Deep Think site ↗
- Current version
- Gemini 3.1 Deep Think
- Best for
- Difficult math, science, engineering, and coding
- Access
- Eligible Google AI Ultra and Ultra for Business plans
- Starting consumer price
- $99.99/month in the U.S.
- App context window
- 1 million tokens on AI Pro and Ultra
- Typical response time
- Several minutes
- Status
- Experimental; limits and availability can change
Overview
What Gemini 3.1 Deep Think is
Deep Think is Google's most compute-intensive reasoning mode for Gemini. The feature first appeared with Gemini 2.5 and Gemini 3, and the current version is built on Gemini 3.1 Pro. Instead of optimizing for a fast conversational response, it explores a problem more broadly and reasons for longer before producing an answer, which can take several minutes.
The public Gemini-app experience is aimed at difficult mathematics, scientific analysis, engineering design, algorithms, and other problems where additional reasoning time can improve a result. It is available only to eligible Google AI Ultra subscribers or organizations with an eligible Ultra for Business license, and users must be at least 18.
Google DeepMind has also used advanced Deep Think systems inside specialized research agents such as Aletheia, which generate, verify, revise, browse literature, and sometimes admit failure. Those research results are important evidence of the underlying approach, but they should not be treated as a promise that every consumer Deep Think prompt runs the same agent workflow or produces publication-ready research.
Use cases
Who Gemini 3.1 Deep Think is best for
The strongest fit depends on the job you need the product to complete, not the size of its feature list.
Scientific analysis
Reason across literature, experimental data, hypotheses, and technical constraints with expert verification.
Engineering design
Evaluate tradeoffs, optimize complex systems, and work through rapid-prototyping questions that benefit from multiple candidate approaches.
Algorithms and competitive programming
Spend more inference time on decomposition, correctness, complexity, and edge cases for difficult coding problems.
High-stakes second opinions
Stress-test a proposed argument or design by asking for both proof and refutation, while keeping a qualified human in charge.
Capabilities
Core Gemini 3.1 Deep Think features
Maximum parallel reasoning
Uses Google's highest Gemini thinking level to explore more candidate paths than Standard or Extended thinking.
Gemini 3.1 Pro foundation
Builds on Google's current high-capability multimodal Pro model for text, files, images, video, math, and code.
Long-context analysis
Google AI Ultra provides a 1-million-token Gemini-app context window for large documents, files, and extended conversations.
Multimodal problem solving
Can reason over supported uploaded text, diagrams, images, files, and video through the Gemini app.
Background completion notifications
Because responses can take minutes, Gemini lets users leave the conversation and notifies them when the result is ready.
Research-oriented performance
Google reports strong results across academic reasoning, mathematics, physics, chemistry, multimodal understanding, and algorithms benchmarks.
Google ecosystem access
Deep Think sits inside an Ultra bundle that also includes higher Gemini limits and other Google AI and storage benefits.
Process
How the Gemini 3.1 Deep Think workflow works
Step 1
Confirm eligibility
Use a personal Google AI Ultra account or an eligible Ultra for Business license, sign in, and confirm regional availability.
Step 2
Select Pro and Deep Think
In Gemini, choose the Pro model, reopen the thinking-level control, and select Deep Think.
Step 3
Frame a verifiable problem
Provide the exact objective, assumptions, data, constraints, acceptable sources, and the form of evidence required.
Step 4
Request adversarial checking
Ask for competing approaches, proof and refutation, edge cases, uncertainty, and checks that would falsify the conclusion.
Step 5
Wait for extended reasoning
Let the multi-minute response complete; Gemini can notify you while you work in another conversation.
Step 6
Verify outside the model
Check citations, calculations, code, experimental claims, and domain assumptions with primary sources and qualified experts before acting.
Cost
Gemini 3.1 Deep Think pricing and free plan
Deep Think is not included with Free, AI Plus, or AI Pro. In the U.S., it is available to eligible Google AI Ultra subscribers. Ultra currently starts at $99.99 per month for 5x AI Pro usage, with a $199.99 per month tier for 20x usage. Exact limits are compute-based and can change; Ultra for Business pricing depends on the organization's Google agreement.
Free / AI Plus / AI Pro
$0 to $19.99/month
These plans provide Gemini access but do not include Deep Think.
- Standard and Extended thinking may be available
- AI Pro is $19.99/month in the U.S.
- 1M-token app context on AI Pro
- Upgrade to eligible Ultra for Deep Think
Google AI Ultra 5x
$99.99/month
The lower-priced U.S. Ultra tier with eligible access to Deep Think.
- 5x AI Pro usage limits
- Starting at 20 TB of cloud storage
- Higher access across Gemini and bundled Google AI products
- Feature and regional eligibility still apply
Google AI Ultra 20x
$199.99/month
The higher-use U.S. Ultra tier for heavier Gemini workloads.
- 20x AI Pro usage limits
- 30 TB storage listed in Google support
- Higher quotas across eligible AI products
- Compute-based caps and weekly limits still apply
Google AI Ultra for Business
Organization pricing
Work or school access depends on an eligible business license and administrator configuration.
- Contact Google or your Workspace administrator
- Availability varies by contract and region
- Enterprise data controls depend on the selected service terms
Pricing checked . Check current pricing at the source ↗
Assessment
Gemini 3.1 Deep Think strengths and limitations
Where it stands out
- Allocates substantially more reasoning effort to difficult problems than normal Gemini chat modes.
- Combines advanced reasoning with multimodal inputs and a large Gemini-app context window.
- Especially strong fit for mathematics, algorithms, scientific analysis, and engineering tradeoffs.
- Background notifications make multi-minute responses less disruptive.
- Research work around Deep Think demonstrates useful generate-verify-revise and human-AI collaboration patterns.
- Included within a broader Ultra bundle rather than sold as a separate per-prompt add-on.
What to consider
- Deep Think is expensive compared with standard consumer AI plans and is restricted to eligible Ultra access.
- Responses can take several minutes and consume limits much faster than lower thinking levels.
- Google describes the feature as experimental and may change, suspend, throttle, or discontinue it without notice.
- Usage caps are compute-based rather than a stable published number of prompts, which makes capacity planning difficult.
- Vendor-reported benchmark and research results do not guarantee correctness on a user's real scientific problem.
- The public Gemini mode should not be confused with Google's specialized Aletheia research-agent workflow.
- Generated proofs, citations, calculations, and scientific conclusions still require independent expert verification.
Compare
Gemini 3.1 Deep Think alternatives
The right alternative depends on the specific output, workflow, controls and budget your project requires.
Consumer
Gemini 3.1 Pro
The underlying Pro model is faster and more broadly accessible for complex work that does not justify maximum Deep Think compute.
Explore Gemini 3.1 Pro →Consumer
Gemini Deep Research Agent
Better suited to long-running web research, source gathering, and synthesized reports rather than proof-style parallel reasoning.
Explore Gemini Deep Research Agent →Science
Claude Science
A scientific-research workspace alternative focused on literature, data, and domain workflows.
Explore Claude Science →Business Operations
ChatGPT
A broader general assistant with reasoning, research, coding, and data-analysis modes at multiple consumer price points.
Explore ChatGPT →Questions
Gemini 3.1 Deep Think FAQs
What is Gemini Deep Think?
Deep Think is the Gemini app's highest thinking level. It spends more time and compute exploring a difficult problem in parallel before responding.
Is Gemini 3 Deep Think now Gemini 3.1 Deep Think?
Yes. The current Google DeepMind product page identifies the mode as Gemini 3.1 Deep Think, built on Gemini 3.1 Pro. The older Gemini 3 name describes an earlier version of the same evolving feature.
How do I access Deep Think?
You need an eligible Google AI Ultra subscription or Ultra for Business license, must be at least 18, and must be in a supported region. In Gemini, select Pro and then choose Deep Think under Thinking level.
How much does Deep Think cost?
In the U.S., Google AI Ultra currently starts at $99.99 per month for 5x AI Pro usage. The 20x Ultra tier costs $199.99 per month. Prices, bundles, and availability vary by country.
How long does a Deep Think answer take?
Google says queries generally take a few minutes. The exact time depends on the prompt and current service capacity.
Is Deep Think available through the Gemini API?
Google's current consumer documentation describes Deep Think inside the Gemini app and eligible Ultra plans. Developers should not assume the app mode is exposed as a generally available API model unless the current Gemini API documentation explicitly lists it.
Can Deep Think do original scientific research?
Google has reported research contributions from advanced Deep Think-based systems working with experts. That does not make every answer novel or correct; research claims, proofs, and citations need independent expert validation.
What is the difference between Deep Think and Deep Research?
Deep Think spends more inference effort reasoning through a difficult problem. Deep Research is designed to search, gather, and synthesize information into a sourced report. The two workflows can complement each other.
Bottom line
Our Gemini 3.1 Deep Think verdict
Gemini 3.1 Deep Think is a premium specialist mode, not the best default for everyday chat. Its value appears when a hard mathematical, scientific, engineering, or algorithmic problem genuinely benefits from multiple reasoning paths and several minutes of compute. Ultra's price and variable limits make it most sensible for researchers and technical teams that can verify the output and use the broader bundle—not for anyone seeking guaranteed answers from a single prompt.
Visit Gemini 3.1 Deep Think website ↗Similar Science Tools
View all Science AI tools →
Nomos 1
Nomos 1 - Nous Research’s powerful open-source AI math reasoning system
TRIBE v2
TRIBE v2 - Meta's predictive foundation model that simulates human brain responses to sights, sounds, and language
SciSpace BioMed Agent
SciSpace BioMed Agent - A specialized biomedical research co-scientist for multi-omics, clinical phenotypes, and lab protocol reasoning
Leanstral 1.5
Leanstral 1.5 - Mistral's open model for writing verified math proofs
AI training for the future of work.
Get access to all our AI courses, hundreds of real-world AI use cases, live expert-led workshops, an exclusive network of AI early adopters, and more.
AI Courses
Get unlimited access to all of our current & upcoming industry-specific AI courses for the duration of your subscription.
Daily Guides
To keep up with the rapid pace of AI, our team publishes AI implementation guides daily. Our library contains 300+ practical use cases to automate real-world work.
Workshops
Join weekly, live, interactive sessions with industry leaders who are at the forefront of AI for hands-on implementation guidance and exclusive insights.
Community
Network with an exclusive community of AI-first professionals who are working smarter with AI. Learn how early adopters are using AI in their work and businesses.