The aim of this paper is to give an upper bound for the dimension of a torus T which acts on a GKM manifold M effectively. In order to do that, we introduce a free abelian group of finite rank, denoted by A(Γ, α, ∇), from an (abstract) (m, n)-type GKM graph (Γ, α, ∇). Here, an (m, n)-type GKM graph is the GKM graph induced from a 2m-dimensional GKM manifold M 2m with an effective n-dimensional torus T n -action which preserves the almost complex structure, say (M 2m , T n ). Then it is shown that A(Γ, α, ∇) has rank ℓ(> n) if and only if there exists an (m, ℓ)-type GKM graph (Γ, α, ∇) which is an extension of (Γ, α, ∇). Using this combinatorial necessarily and sufficient condition, we prove that the rank of A(Γ M , α M , ∇ M ) for the GKM graph (Γ M , α M , ∇ M ) induced from (M 2m , T n ) gives an upper bound for the dimension of a torus which can act on M 2m effectively. As one of applications of this result, we compute the rank of A(Γ, α, ∇) of the complex Grassmannian of 2-planes G 2 (C n+2 ) with some effective T n+1 -action, and prove that the T n+1 -action on G 2 (C n+2 ) is the maximal effective torus action which preserves the standard complex structure.Theorem 1.2. Let (Γ, α, ∇) be an abstract (m, n)-type GKM graph. Then, the following two statements are equivalent:(1) rk A(Γ, α, ∇) ≥ ℓ for some n ≤ ℓ ≤ m;(2) there is an (m, ℓ)-type GKM graph (Γ, α, ∇) which is an extension of (Γ, α, ∇).Because a GKM manifold (M 2m , T n ) defines an (m, n)-type GKM graph (see Section 4), Theorem 1.2 implies that the maximal dimension of torus which can act on a GKM manifold M is bounded from above by the rank of the group of axial functions of the GKM graph induced from M . Namely, we obtain the main result of this paper as follows (see Section 4 for details):Then, the T n -action on M 2m does not extend to any T ℓ+1 -action preserving the given almost complex structure.Hence, if rk A(Γ M , α M , ∇ M ) = n, then the T n -action on M 2m is maximal among torus actions which preserve the given almost complex structure.Remark 1.4. Shunji Takuma also obtains a partial answer to Problem 1.1 by introducing an obstruction class for the extension of an (m, n)-type GKM graph to an (m, n + 1)-type GKM graph in his note [19]. Theorem 1.2 may be regarded as the generalization of his result. Problem 1.1 is reminiscent of the computation of the torus degree of symmetry of a manifold X (see [12]), i.e., the maximal dimension of a torus which can act on X effectively. A torus degree of symmetry has been studied for many classes of manifolds, in particular from differential geometry (see e.g. [4,12,21,23]). Corollary 1.3 may be regarded as to give an upper bound of the torus degree of symmetry of an invariant almost complex structure of a GKM manifold. As an application of Corollary 1.3, in the final section (Section 5), we compute the torus degree of such symmetry for the complex Grassmannian of 2-planes, denoted as G 2 (C n+2 ) ≃ GL(n + 2, C)/GL(2, C) × GL(n, C) ≃ U (n + 2)/U (2) × U (n). Namely, we compute rk A(Γ M , α M , ∇ M ) for ...