LOCK
01
04
06
07
08
09
10
11
12
13
17
18
20
Open-source AI modelsfor local use
Understand open-weight AI models: licences, instruct vs code models, hardware fit, and how curated catalogs make local open-source AI practical.
Open weights vs open source software
“Open” can mean downloadable weights, training code, data, or all three. Always read the licence. Some models are research-only; others allow commercial use under Apache-2.0-style terms.
Popular categories
General instruct models for chat and drafting; code-specialised models for repositories; embedding models for search; multimodal models for vision. Pick by job, not hype alone.
Why catalogs beat scavenger hunts
Random forum links skip hardware floors, licence notes, and purpose. A product catalog documents good-for / not-for so teams ship safely.
Frontier open libraries
Sparse multi-trillion-parameter class open models exist for Lab hardware. They need serious disk and RAM and honest latency expectations. Helmsman-style control planes exist to pilot them — not to pretend they are free chat.
S◉LOCK model plugins
Curated plugins for Home and Lab profiles, installable via CLI on Founding builds, with docs and bindings. First-party S◉LOCK packs are on the roadmap when quality warrants.
Next step: See which open models made the Helmsman catalog, check what they ask for in hardware floors, or request Founding Preview access.
FAQ
- Are open models safe?
- Safety is layered: licence compliance, sandboxing, human review for high-stakes output, and refusing to treat any model as oracle truth.
- Can businesses use open models commercially?
- Often yes under Apache-style licences — but verify each model. Do not assume.
Related guides
Open weights deserve owned metal
Founding Preview is the limited early door to a desktop built around exactly this idea: open models on hardware you own, EU-born, no automatic cloud fallback.
Request Founding access