Google's OA isn't a single test — it's three parallel tracks: Grad OA (New Grad SDE), Intern OA (summer / year-round), and invitation-only Foobar (Google's internal recruiter pool). Each has its own platform, format, and difficulty. Below: 2026 spring-cycle breakdown + prep strategy.
Google OA Three-Track Snapshot
| Track | Platform | Duration | Questions | Difficulty |
|---|---|---|---|---|
| Grad OA | Google in-house | 90 min | 2 | LC Medium-Hard |
| Intern OA | HackerRank | 60 min | 2 | LC Easy-Medium |
| Foobar | Google in-house (terminal sim) | unlimited | 9 in 5 levels | LC Medium → Hard |
Grad OA Problem Lines
Type 1: Graph + State Search
n-node directed graph, each node carries a cooldown. From node 0, find the minimum total time to visit all nodes (re-visit only after cooldown ≥ T).
from functools import lru_cache
def min_visit_all(n, edges, cooldown):
g = [[] for _ in range(n)]
for u, v, w in edges:
g[u].append((v, w))
@lru_cache(maxsize=None)
def dp(node, visited_mask, last_visit_time_tuple):
if visited_mask == (1 << n) - 1:
return 0
best = float('inf')
for v, w in g[node]:
new_mask = visited_mask | (1 << v)
sub = dp(v, new_mask, last_visit_time_tuple)
if sub != float('inf'):
best = min(best, w + sub)
return best
return dp(0, 1, tuple([0] * n))
Prompts typically cap
n ≤ 12, hinting at bitmask DP.
Type 2: DP / Interval / Number Theory
Classics like LC 1547 / 312 / 1392. Google favors plain wording + non-trivial state compression.
Intern OA Problem Lines
Smaller volume, faster pace: 2 problems in 60 minutes. Typical combos:
- 1 string + 1 tree / graph
- 1 sliding window + 1 binary search
Example: String Reorganization
import heapq
from collections import Counter
def reorganize_string(s):
cnt = Counter(s)
if max(cnt.values()) > (len(s) + 1) // 2:
return ""
h = [(-v, k) for k, v in cnt.items()]
heapq.heapify(h)
res = []
prev = (0, '')
while h:
v, k = heapq.heappop(h)
res.append(k)
if prev[0] < 0:
heapq.heappush(h, prev)
prev = (v + 1, k)
return ''.join(res)
Foobar Problem Lines
Foobar is 5 levels, 9 problems, progressing LC Medium → Hard. Only Java / Python accepted; no auto-test feedback — one submission, pass/fail.
- Level 1: 1 problem (warmup)
- Level 2: 2 problems
- Level 3: 3 problems
- Level 4: 2 problems
- Level 5: 1 problem (Hard, often number theory / bitmask / competitive)
Completing Level 3 triggers a Google recruiter reach-out; Level 5 = onsite direct.
Office-by-Office Differences
| Office | Preferred Topics |
|---|---|
| Mountain View | Search systems, ad ranking, recommendation |
| New York | Finance / ads / Cloud |
| Zurich | Compiler / number theory / large-scale distributed |
| Seattle | Cloud / SRE / Storage |
7-Day Sprint
| Day | Task |
|---|---|
| D1 | Graph: topo, Dijkstra, Floyd, Tarjan |
| D2 | DP: knapsack, interval, bitmask, probability |
| D3 | Strings: KMP, Trie, Z-function |
| D4 | LC Google-tagged Medium ×20 |
| D5 | LC Google-tagged Hard ×10 |
| D6 | Timed 90-min Grad OA mock |
| D7 | Debrief + gap drills |
FAQ
Grad OA prompt language?
English primarily. Some China-office Intern OAs offer Chinese prompts.
How does a Foobar invite happen?
Mainly Google pushes invites based on search behavior + GitHub activity. A minority enter via GCJ / Kickstart.
Pass rate?
Community reports: Grad ~20%, Intern ~35%, Foobar Level 3 = recruiter reach-out.
Result timeline?
Usually 7–21 days. Grad OA pass → phone screen; Intern OA → host matching.
Preparing Google OA / VO?
We were glad to help this cohort pass Google Grad / Intern OA. Many candidates told us LC Google-tag grinding triggers "I can't do Hard" anxiety, but 80% of real Google OAs are "plain wording + bitmask DP" combos — the key is spotting the hidden state.
If you're prepping Google, Meta, DeepMind, or Waymo OA / VO and feel directionless or unsteady on pacing, contact oavoservice. We tailor OA assistance to your gaps and lock in the Grad / Intern / Foobar three-track patterns.
👉 Add WeChat: Coding0201 — grab the Google prep pack.
Contact
Email: [email protected]
Telegram: @OAVOProxy