Back to the interactive map
Inference Workload

Code Completion

Streaming token prediction tuned on code — fill-in-the-middle (FIM), whole-file generation, and repo-level reasoning.