Select runtimes to compare side by side. Click chips below to toggle selection.
| Metric | CUDA Runtime | llama.cpp | vLLM |
|---|---|---|---|
| Score | B-65 | C+63 | F35 |
| Type | |||
| Execution | aot | aot | jit |
| Interface | sdk | cli | api |
| Cold Start | 100ms | 100ms | 5000ms |
| Memory | 500MB | 50MB | 2000MB |
| Startup | 50ms | 10ms | 3000ms |
| Isolation | hardware | process | process |
| Maturity | production | production | production |
| Languages | C, C++, Python | C, C++ | Python |
| License | Proprietary | MIT | Apache-2.0 |
| Links |