86–113x faster — zerodep strips comments via lightweight regex then delegates to C-accelerated json.loads, while commentjson builds a full Lark LALR parse tree.
Speedup ratio improves with data size, showing lower per-element overhead.
On par with ndjson — both use the same batch strategy: join lines into a JSON array and parse in a single json.loads call, minimizing Python-level loop overhead.
3.6–4.1x faster than jsonlines — jsonlines uses a per-line json.loads loop with a Reader wrapper, adding significant Python overhead.
JSONC bonus — unlike ndjson/jsonlines, zerodep automatically falls back to per-line JSONC processing when comments or trailing commas are present. Clean JSONL gets the fast path; commented JSONL still works.