Who it's for

DoesItLocal is for people and software making the local-vs-frontier routing decision — repeatedly, under a cost or privacy constraint, who can’t afford to guess wrong.

Primary: the cost-conscious local-LLM developer

Someone running Ollama or LM Studio on a MacBook (or a workstation) who wants to offload as much work as possible to a free local model without silently degrading quality. They already suspect a 27B local model can summarize, extract, and draft — but they don’t trust it with the agentic coding loop, and they want a defensible, current answer for each task instead of re-litigating it by feel. They come to DoesItLocal to ask “is this task safe for the model I have, and if not, what’s the cheapest thing that is?”

Primary: coding agents and routers

The biggest consumer is software. A router (LiteLLM, a FrugalGPT-style cascade, Gemini CLI’s local Gemma classifier) needs to decide, per request, whether to spend a local-model call or a frontier-model call. DoesItLocal exposes the per-task verdict as a typed API/MCP response so the router can shortlist locally and escalate only when the task isn’t safe for local — turning “guess and hope” into a lookup backed by evidence and a verifier. This is the audience that makes the data worth keeping fresh.

Secondary: the privacy-bound user

Some tasks must stay on the machine — proprietary code, personal data, anything under an NDA. For this user “local” isn’t about saving tokens; it’s the only acceptable option, and the question flips: “this has to run locally — is it actually safe to, or do I need a verification step / a different local model / to redesign the task?” DoesItLocal flags privacy-bound tasks where local is the safer choice even when a frontier model would score higher.

Secondary: the team standardizing a routing policy

A small team that wants a shared, written policy for “what we run locally vs. what we pay for,” instead of each engineer deciding ad hoc. DoesItLocal is the reference they point at — and the place they contribute their own findings back.

Not the audience

• People who just want the single best model. That’s a leaderboard’s job; DoesItLocal is about task-level fitness, not crowning a winner.
• Pure benchmark researchers. The verdicts lean on evals but the product is an applied routing reference, not a benchmark suite.
• Anyone wanting DoesItLocal to run the model for them. It publishes facts about tasks; it is not an inference host or an installable router.