Select runtimes to compare side by side. Click chips below to toggle selection.
| Metric | Metal | Python (CPython) | ExLlamaV2 | KoboldCpp |
|---|---|---|---|---|
| Score | B74 | B71 | D47 | F38 |
| Type | Language | |||
| Execution | aot | interpreted | aot | hybrid |
| Interface | sdk | cli | sdk | gui |
| Cold Start | 50ms | 50ms | 1000ms | 1500ms |
| Memory | 100MB | 15MB | 300MB | 400MB |
| Startup | 20ms | 10ms | 200ms | 300ms |
| Isolation | hardware | process | process | process |
| Maturity | production | production | stable | stable |
| Languages | Swift, Objective-C, C++ | Python | Python, C++, CUDA | C++, Python |
| License | Proprietary | Other | MIT | AGPL-3.0 |
| Links |