How does the TCP/IP model compare to the OSI model?

Both models describe how networked communication is layered, but they come from different worlds: OSI was a standards-body design meant to be protocol-neutral, while TCP/IP grew from the protocols that actually run the internet. Knowing how they line up shows you understand theory and practice.

The TCP/IP four layers

1. Link (Network Access) — physical transmission plus framing and MAC addressing on the local network. Combines OSI Physical + Data Link.
2. Internet — logical addressing and routing with IP, ICMP, and friends. Maps to OSI Network.
3. Transport — TCP and UDP: end-to-end delivery, ports, reliability. Maps to OSI Transport.
4. Application — everything an app does: HTTP, DNS, TLS, SMTP. Absorbs OSI's Session, Presentation, and Application layers.

How they map

The Transport and Internet/Network layers line up almost one-to-one. The divergence is at the edges: TCP/IP folds OSI's three top layers into one Application layer (because in practice an application like a browser handles session, encoding, and protocol logic together) and folds OSI's two bottom layers into one Link layer.

Which to use when

Reach for OSI when teaching, troubleshooting, or talking about where a control sits — its finer granularity ("this is a Layer 4 firewall, that's a Layer 7 proxy") is more precise. Reach for TCP/IP when discussing how real traffic flows, because those are the protocols on the wire. The internet standardized on TCP/IP largely because working, deployed implementations beat OSI's slower, committee-driven protocol suite.

Interviewers want the four layer names, a correct mapping (especially that Application covers OSI's top three), and an explanation of why TCP/IP is the practical model.