Everyone else picks the model first. This starts with the machine.

$ llmfit

Hundreds of models and providers, and one command that returns the subset your hardware can actually hold. The catalog is the variable here. Your machine is the constant, and it decides where the line falls.

The catalog, sorted by what a machine can hold

bed widths are decorative, no model named

too large

—

your hardware The line that sorts the catalog into runs and does not run. Nothing in the description says where it sits, because it sits somewhere different for every reader.

the input
not published

fits

what the command returnsthe beds below the line no model, provider, or VRAM figure published

One command is the whole interface

No account, no config file, no model picker. A single invocation that reads the machine and answers, which is the smallest surface a tool like this could have.

Rust for a hardware probe

Reading memory, GPU, and driver details across three operating systems is the kind of job where a single static binary saves more trouble than the language choice costs.

What is not on this page

No model names, no provider list, no memory thresholds, no quantisation. The trending description is two sentences and names none of it.

Concept mock, not affiliated with the project. The complete GitHub Trending description on 10 September 2026 is "Hundreds of models and providers. One command to find what runs on your hardware." The seven beds are drawn as anonymous bands to show the sorting the command performs. No model, provider, memory figure, or threshold appears here, because that sentence publishes none.