#include <iostream>
#include <cstdint>
#include <vector>
#include <algorithm>

using namespace std;

static inline constexpr vector<int64_t> prepare()
{
	vector<bool> is_prime(10'000'000, true);
	is_prime[0] = is_prime[1] = false;
	for (uint32_t i = 2; i != 10'000'000; ++i)
		if (is_prime[i] && static_cast<uint64_t>(i) * i < 10'000'000)
			for (uint32_t j = i * i; j < 10'000'000; j += i)
				is_prime[j] = false;

	vector<int64_t> candidates;
	candidates.reserve(10'000'000);
	for (uint32_t i = 9'999'999; i != 3; i -= 2)
		if (is_prime[i] && is_prime[i - 2])
			candidates.push_back(static_cast<int64_t>(i - 2) * i);
	candidates.push_back(-1);

	return candidates;
}

static inline constexpr vector<int64_t> solve(const uint32_t& T, const vector<int64_t>& N)
{
	const vector<int64_t>& candidates = prepare();
	vector<int64_t> ans(T);
	for (uint32_t i = 0; i != T; ++i)
		ans[i] = *lower_bound(candidates.begin(), candidates.end(), N[i], greater<int64_t>());

	return ans;
}

static inline void output(const uint32_t T, const vector<int64_t>& ans)
{
	for (uint32_t i = 0; i != T; ++i)
		cout << ans[i] << '\n';
}

int main()
{
	cin.tie(nullptr);
	ios::sync_with_stdio(false);
	
	uint32_t T, i;
	cin >> T;
	vector<int64_t> N(T);
	for (i = 0; i != T; ++i) cin >> N[i];

	output(T, solve(T, N));
	return 0;
}