Open Weight Models
A curated, actively maintained list of open-weight AI models selected for commercially exploitable licenses, no geographic restrictions, and verified benchmarks, organized by category and VRAM tier.
Open Weight Models is a curated list of open-weight AI models chosen against explicit selection criteria: a commercially exploitable license with no geographic restriction, and a VRAM footprint of 128 GB or less when quantized to Q4 for its main tier, with an extended tier for larger models. It was built to decide which models to support in a separate local LLM inference engine written in Rust and hand-written assembly.
Entries are grouped by category, generalists, code models, compact and edge models, reasoning and math models, vision and multimodal models, alternative architectures and web world models, and the list flags models whose license carries a revenue or monthly-active-user cap, warning that these are not interchangeable with a permissive license such as Apache 2.0 or MIT despite being listed for completeness.
It is updated periodically with dated addition notes, most recently in mid-2026.