~ / local

Running models on your own hardware

Open weights are free. The machine to run them is not. This section answers the five questions that decide whether self-hosting is worth it for you, with the numbers in tables rather than spread across a hundred forum replies.

Coding

Which GPU runs a coding model, what the ceiling is against Claude and GPT, why open weights often still need a server, and the cheapest hardware that actually works.

live

8-12 GB cards

The playbook for production work on an ordinary GPU: which models fit, six deploy steps, and the honest pros and cons against a subscription.

live

Image

Local image generation: model sizes, VRAM per resolution, throughput, and what a workstation costs against a subscription.

planned

Hardware and video

MTT S5000, Mac Studio Ultra 512GB and RTX Pro 6000: prices with source labels, and a VRAM fit matrix for LTX-2.5, HunyuanVideo 1.5 and Wan.

live

Music and speech

Local audio generation and text to speech: model choices, memory, and real-time factors.

planned
Why this section exists. The people selling API access have no reason to explain what you could run yourself, and the people who do run models locally mostly write it up in scattered threads where the hardware, the quantisation and the quality claims never appear in the same place. Every page here keeps those together, states the price with the date it was checked, and says plainly when a number has never been measured rather than filling the gap with a guess.

Start with coding Local models and hardware rows