PortPeek Logo PortPeek v0.2.0
GitHub Download
11434 TCP
Local AI & Large Language Models Unassigned / De-facto Local AI Standard

Port 11434 — Ollama (Local LLM Engine & API)

The default HTTP REST API endpoint for Ollama, enabling local execution of open-source Large Language Models (Llama 3, Mistral, DeepSeek, Gemma 2, Phi-3, Qwen) on Windows GPU and CPU.

Default Protocol TCP / IP
Typical Localhost URL http://localhost:11434
Default Process ollama.exe / ollama_app.exe
Primary Framework Ollama

1. What is Port 11434?

Port 11434 is the standard port for Ollama, the tool that revolutionized local AI by packaging quantization, GPU acceleration (CUDA, ROCm, Vulkan), model weights, and REST APIs into a single executable.

The Ollama engineering team selected 11434 as a clean, unassigned port in the user registered space. The number has quickly become the recognized universal standard for local AI inference across developer tooling, web UIs (Open WebUI, Enchanted), and code editor extensions (Continue.dev, Cursor).

2. How Port 11434 Works Under the Hood

1
Socket Creation & Binding

Process creates a Winsock socket (`AF_INET`/`AF_INET6`, `SOCK_STREAM`) and binds to local port 11434.

2
Kernel TCP Listener Table

Windows TCP/IP stack records the listening state in `MIB_TCPTABLE_OWNER_PID`.

3
Request Handling & Loopback Dispatch

Inbound HTTP/TCP traffic to `127.0.0.1:11434` is delivered directly through Windows loopback with zero physical NIC overhead.

When Ollama runs on Windows: 1. `ollama.exe` (or `ollama_app.exe` tray daemon) binds a Go-based HTTP server to `127.0.0.1:11434`. 2. When an inference request is posted to `/api/generate` or `/v1/chat/completions`, Ollama loads the quantized model (GGUF format) into VRAM via `llama.cpp` C++ runner subprocesses. 3. Response tokens are streamed back in real-time over the HTTP connection as Server-Sent Events (SSE) or JSON chunks. 4. If no requests arrive within the keep-alive timeout (default 5 minutes), Ollama unloads model weights from GPU VRAM to preserve system memory while maintaining the listening socket on 11434.

3. Common Conflicts & 'Error: listen tcp 127.0.0.1:11434: bind: address already in use'

⚠️ Common Error Message
Error: listen tcp 127.0.0.1:11434: bind: address already in use

On Windows, the Ollama GUI installer configures Ollama to run automatically on Windows startup as a background system tray process (`ollama_app.exe`). If you subsequently open a terminal and manually type `ollama serve` or run an Ollama Docker container mapping `11434:11434`, you will get: `Error: listen tcp 127.0.0.1:11434: bind: address already in use` Knowing that the background Ollama daemon is already happily listening on port 11434 means you can immediately start executing prompts against `http://localhost:11434` without running `serve` manually.

💡 Pro Tip: Ollama exposes an OpenAI-compatible REST API at `http://localhost:11434/v1` allowing zero-code drops into LangChain, LlamaIndex, and Copilot tools.

4. How to Find & Stop the Process on Port 11434

If port 11434 is locked by an orphaned process or background service, choose one of the following methods:

Terminal Commands (Manual) PortPeek (1-Click Instant)

PowerShell (Recommended):

Get-Process -Id (Get-NetTCPConnection -LocalPort 11434).OwningProcess | Stop-Process -Force

Classic CMD (Command Prompt):

for /f "tokens=5" %a in ('netstat -aon ^| findstr :11434') do taskkill /f /pid %a
PortPeek

Or simply use PortPeek

Never memorize netstat flags or parse PIDs again.

1. Press Win + Alt + P anywhere on Windows.
2. See Port 11434 with its owning process (ollama.exe / ollama_app.exe) and memory usage.
3. Click Stop Process to kill it instantly, or click the item to launch in your browser.
Download PortPeek (Free & Open Source)

5. Standards & Related Ports

Related Ports:

Port 8000Port 7860Port 8501Port 8001

Official Specifications:

  • IANA Service Name and Transport Protocol Port Number Registry
  • RFC 7230, OpenAI Compatible REST API Standard
  • Windows Sockets 2 (Winsock) API Documentation