Select runtimes to compare side by side. Click chips below to toggle selection.
| Metric | Python (CPython) | Candle | ExLlamaV2 | KoboldCpp |
|---|---|---|---|---|
| Score | B71 | D48 | D47 | F38 |
| Type | Language | |||
| Execution | interpreted | jit | aot | hybrid |
| Interface | cli | sdk | sdk | gui |
| Cold Start | 50ms | 300ms | 1000ms | 1500ms |
| Memory | 15MB | 200MB | 300MB | 400MB |
| Startup | 10ms | 50ms | 200ms | 300ms |
| Isolation | process | process | process | process |
| Maturity | production | stable | stable | stable |
| Languages | Python | Rust | Python, C++, CUDA | C++, Python |
| License | Other | Apache-2.0 | MIT | AGPL-3.0 |
| Links |