Why L4 at Google is different from L4-equivalent at peers
Google L4 maps roughly to Meta E4, Amazon SDE2, Microsoft Level 61 — but the loop is calibrated differently, and prepping the wrong rubric is the most common over-correction we see.
The biggest difference: Google L4 typically does not include a system design round. Per multiple 2025 candidate reports (hellointerview.com L4 guide, IGotAnOffer L4 guide, Onsites.fyi 2025 L4 guide), the L4 onsite focuses on coding rounds plus a Googleyness/behavioral round. System design is reserved for L5+. Meta E4 candidates always face system design; Amazon SDE2 always faces the LP-driven Bar Raiser. Google L4 weights coding output more heavily than peers — it's the primary signal that distinguishes L4 from a down-level to L3.
Google also evaluates General Cognitive Ability (GCA) as a meta-signal across rounds: how you adapt when a problem twists, whether you ask clarifying questions before pattern-matching, whether you can articulate trade-offs without prompting. GCA isn't a separate round — it's the dimension being scored across all rounds. Strong rote LeetCode performance with weak GCA signals can still result in a down-level.