Llama.cpp

I tested Speculative decoding (Multi-Token Prediction, MTP) performance in Qwen 3.6 27B and 35B on an RTX 4080 with 16 GB VRAM.

llama.cpp router mode is one of the most useful changes to llama-server in years. It finally gives local LLM operators something close to the model management experience people expect from Ollama, while keeping the raw performance and low-level control that make llama.cpp worth using in the first place.

This page is a practical reference for agentic LLM inference tuning (temperature, top_p, top_k, penalties, and how they interact in multi-step and tool-heavy workflows).

For a long time, llama.cpp had a glaring limitation:
you could only serve one model per process, and switching meant a restart.

Here I am comparing speed of several LLMs running on GPU with 16GB of VRAM, and choosing the best one for self-hosting.

Soon you are juggling vLLM, llama.cpp, and more—each stack on its own port. Everything downstream still wants one /v1 base URL; otherwise you keep shuffling ports, profiles, and one-off scripts. llama-swap is the /v1 proxy before those stacks.

I have tested how OpenCode works with several locally hosted on Ollama and llama.cpp LLMs, and for comparison added some Free models from OpenCode Zen.

llama.cpp is a C/C++ inference engine for GGUF models: llama-cli for interactive chat, llama-completion for scripted prompts, and llama-server for an OpenAI-compatible HTTP API.

Strategic guide to hosting large language models locally with Ollama, llama.cpp, vLLM, or in the cloud. Compare tools, performance trade-offs, and cost considerations.