Nano GPT logo
NanoGPT
Nano GPT logo
NanoGPT
Conversations
Create Media
Workspace
Gallery
Models
Pricing

Account

Balance
Usage
Settings
API
Presets
Invitations
Subscription
Teams

Resources

Benchmarks
Suggestions and bugs
Batch API
Blog
Help
Applications
Earn
Partners
Terms

|

Privacy

|

Refunds

Private AI

Back to Blog

Web Access for All Models with LinkUp

Mar 2, 2025
Web Access for All Models with LinkUp

Web Access for All Models: Making Every AI Smarter

We're thrilled to announce that we've added web search capabilities to all our models through our partnership with LinkUp! This represents a significant upgrade to our platform, giving every AI model the ability to tap into the internet for real-time information.

Why This Matters

A large language model without internet access is like a brilliant person who's been isolated in a cave for a year - they don't know about recent events, the latest products, or breaking news. By connecting our models to the internet through LinkUp, we solve this fundamental limitation, allowing our AI models to generate responses that are:

  • More accurate and factual
  • Updated with current information
  • Better able to reference specific details
  • Capable of citing sources

How It Works

We've made enabling web access incredibly simple:

  • On Desktop: Click "enable web" next to any model
  • On Mobile: Tap the web icon
  • Via API: Append :online to any model name

Once enabled, the model gains access to 10 web search results related to your query, dramatically improving its ability to provide accurate, up-to-date information.

Search Depth Options

You can choose between two search depths based on your needs:

  • Standard Search: Returns 10 results quickly, perfect for straightforward questions
  • Deep Search: Continues searching iteratively if the initial results aren't sufficient, yielding more comprehensive information

Via API, you can enable deep search by appending :online/linkup-deep to the model name.

Performance Benchmarks

LinkUp Performance Benchmarks

LinkUp has achieved state-of-the-art performance on OpenAI's SimpleQA benchmark, scoring an impressive 90.10%. For comparison:

  • LinkUp Deep Search: 90.10%
  • Exa: 90.04%
  • Perplexity Sonar Pro: 86%
  • LinkUp Standard Search: 85%
  • Perplexity Sonar: 77%
  • Tavily: 73%
  • Grok 3 (without web access): 45%

These results demonstrate that internet connectivity is arguably more important than model size when it comes to factuality. By connecting GPT-4o-mini to the web, they've seen a 10x improvement in factuality, making it twice as accurate as models like o1, which have superior performance in other areas.

How LinkUp Does It

LinkUp delivers exceptional performance through:

  1. Advanced Search Algorithms: Their proprietary technology ensures more relevant and accurate results
  2. Data Integrations: Native integration with various data sources for high-quality information
  3. Real-time Processing: Purpose-built infrastructure that can surface information published less than a minute ago

Pricing

We've kept pricing simple and affordable:

  • Standard Web Search: $0.006 per query
  • Deep Web Search: $0.06 per query

Note that enabling web search increases the input token count, which may affect the total cost of your queries.

What's Next

We'd love to hear how you're using web-enabled models! Share your experiences and feedback with us through our usual channels.

Happy searching!

Milan de Reede

Milan de Reede

CEO & Co-Founder

milan@nano-gpt.com

Related articles

Continue with more NanoGPT guides and research on this topic.

Connecting AI Models to Skype for Business

Integrate conversational and image AI into Skype for Business to automate workflows, secure data locally, and enable real-time web searches.

Apr 10, 2026

Grid Search vs. Random Search for Image Models

Explore the differences between grid search and random search for hyperparameter tuning in image models, and find which method suits your needs best.

Nov 11, 2025

Top Tools for Resolving Dependency Conflicts in AI Models

Dependency conflicts break AI projects—use pinning, Conda/Mamba, AI debuggers, and unified model APIs to prevent GPU and runtime failures.

Jun 2, 2026

RAM vs. VRAM: Which Matters for AI Models?

Compare RAM and VRAM for local AI: which limits model size, affects token speed, and hardware tips for running 7B–70B models.

Jan 2, 2026
Back to Blog