For the constant of ellipticity $\nu=\Lambda/\lambda$ close to $1$
for elliptic equation of the general form 
$$\sum_{i,j=1^n}a^{ij}(x)u_{ij}(x)=0$$
Cordes H. O. [*Uber die erste Randwertaufgabe bei quasilinearen Differentialgleichun-
gen zweiter Ordnung in mehr als zwei Variablen, Math. Ann., 1956. V. 131. P. 278—
312*] proved that provided a solution is smooth enough it belongs locally to $C^{1+\beta_0}$ where $\beta_0\in(0,1)$ depends on $\nu$ and the estimate 
$$
\|\partial_{x_i} u\|_{C^{1+\beta_0}(B_{1/2})}\le C(n,\nu)\sup_{B_1}|u|
$$
holds. So at least for smooth solutions the required estimate is true.

From the other hand for general $\nu\in(0,1)$ Safonov M.V. [*Unimprovability of estimates of Hölder constants for solutions of linear elliptic equations with measurable coefficients. (English. Russian original) [J] Math. USSR, Sb. 60, No.1, 269-281 (1988); translation from Mat. Sb., Nov. Ser. 132(174), No.2, 275-288 (1987)*]  proved for $n=3$ that such an estimate is not true. Namely there exists $\nu_0\in(0,1)$ s.t. the estimate above does not holds for any $\nu\in(0,\nu_0]$, $0<\beta<1$, $C>0$.  Though the answer to the question posed is not directly answered by this result of Safonov it makes look like a a possibility that the answer can be negative at least for $\nu<\nu_0$ and $n\ge4$ (where one can consider solutions not depending on $x_n$.)