No open-weight model matches a hosted frontier model outright. But for most real work there is a community-regarded closest open-weight model you can run on your own hardware — free, private, and offline. Pick a frontier model to see the closest local equivalents by capability, and exactly what it takes to run each one.
Anthropic's largest frontier model — strongest at hard coding, agentic and reasoning work.
5 local alternatives →
Anthropic's balanced workhorse — the everyday coding and agentic model.
5 local alternatives →
OpenAI's current flagship closed model.
4 local alternatives →
Google's current flagship multimodal model.
4 local alternatives →
Equivalence is an editorial "closest by capability class" judgement, never a benchmark claim. Sizes, VRAM and tokens/sec are estimates derived from the catalog and the same fit engine used across the site. See also the full model catalog and hardware pages.