diff --git a/src/plugins/intel_cpu/src/nodes/kernels/x64/rms_kernel.cpp b/src/plugins/intel_cpu/src/nodes/kernels/x64/rms_kernel.cpp
index 7778ebb12450f0..d3307bb12fbebd 100644
--- a/src/plugins/intel_cpu/src/nodes/kernels/x64/rms_kernel.cpp
+++ b/src/plugins/intel_cpu/src/nodes/kernels/x64/rms_kernel.cpp
@@ -160,6 +160,7 @@ void jit_rms_kernel<isa>::generate() {
     reduce_vmm_to_scalar(vmm_rsqrt, vmm_sum0, vmm_sum1, vmm_sum3, vec_size);
 
     // mean(x^2)
+    OPENVINO_ASSERT(m_jcp.data_size != 0);
     mov(reg_tmp.cvt32(), float2int(1.0f / m_jcp.data_size));
     vmovd(xmm_tmp, reg_tmp.cvt32());
     vmulss(xmm_rsqrt, xmm_rsqrt, xmm_tmp);
diff --git a/src/plugins/intel_cpu/src/nodes/llm_mlp.cpp b/src/plugins/intel_cpu/src/nodes/llm_mlp.cpp
index b475a602c3cd1a..91a72fdbeb4cab 100644
--- a/src/plugins/intel_cpu/src/nodes/llm_mlp.cpp
+++ b/src/plugins/intel_cpu/src/nodes/llm_mlp.cpp
@@ -37,8 +37,6 @@ class LinearKsplit2 {
 
     LinearKsplit2() {}
 
-    ReduceAdd2bh* p_jit_reduce2bh;
-
     // weight [N, K]
     // Gate & Up are interleaved in N dimension: 16-gate / 16-up
     // and post-ops will compute  silu(gate)*up in unit of 16 elements
@@ -201,7 +199,7 @@ class LinearGateUp {
         bool quantized_int8 = config.gate_up_quantized;
 
         auto reg_blk_K_size = quantized_int8 ? REG_BLK_K_SIZE_I8 : REG_BLK_K_SIZE;
-        auto cache_blk_k_size = quantized_int8 ? CACHE_BLK_K_SIZE : CACHE_BLK_K_SIZE;
+        auto cache_blk_k_size = CACHE_BLK_K_SIZE;
         auto weight_element_size = quantized_int8 ? sizeof(int8_t) : sizeof(ov::float16);
 
         // prepare weights, split N among threads
diff --git a/src/plugins/intel_cpu/src/nodes/qkv_proj.cpp b/src/plugins/intel_cpu/src/nodes/qkv_proj.cpp
index ce7bfae07591d6..a7f0286c6b667c 100644
--- a/src/plugins/intel_cpu/src/nodes/qkv_proj.cpp
+++ b/src/plugins/intel_cpu/src/nodes/qkv_proj.cpp
@@ -75,7 +75,7 @@ struct QKVProjection::Executor : public QKVProjection::ExecutorBase {
         // and activations will be dynamically per-token quantized and using AMX-INT8 to get the result
         bool quantized_int8 = m_node->m_config.quantized;
 
-        auto cache_blk_k_size = quantized_int8 ? CACHE_BLK_K_SIZE : CACHE_BLK_K_SIZE;
+        auto cache_blk_k_size = CACHE_BLK_K_SIZE;
         auto weight_element_size = quantized_int8 ? sizeof(int8_t) : sizeof(ov::float16);
 
         auto K = w0.size(1);
diff --git a/src/plugins/intel_cpu/src/nodes/scaled_attn.cpp b/src/plugins/intel_cpu/src/nodes/scaled_attn.cpp
index 7fe3fc8dc5045d..71d39a1fce5ba8 100644
--- a/src/plugins/intel_cpu/src/nodes/scaled_attn.cpp
+++ b/src/plugins/intel_cpu/src/nodes/scaled_attn.cpp
@@ -109,7 +109,7 @@ struct MHAKernel {
     }
 
     PlainTensor causal_mask;
-    bool select_nfltmax_at_0;  // set attn_score to -FLT_MAX when causal_mask[...] equal to this
+    bool select_nfltmax_at_0 = false;  // set attn_score to -FLT_MAX when causal_mask[...] equal to this
     void set_causal_mask(PlainTensor mask, bool _select_nfltmax_at_0) {
         causal_mask = mask;
         select_nfltmax_at_0 = _select_nfltmax_at_0;
@@ -526,7 +526,7 @@ struct MHAKernel<ScaledDotProductAttention::KT_ACL, T> {
     }
 
     PlainTensor causal_mask;
-    bool select_nfltmax_at_0;  // set attn_score to -FLT_MAX when causal_mask[...] equal to this
+    bool select_nfltmax_at_0 = false;  // set attn_score to -FLT_MAX when causal_mask[...] equal to this
     void set_causal_mask(PlainTensor mask, bool _select_nfltmax_at_0) {
         causal_mask = mask;
         select_nfltmax_at_0 = _select_nfltmax_at_0;
@@ -674,7 +674,7 @@ struct MHAKernel<ScaledDotProductAttention::KT_MLAS, float> {
     }
 
     PlainTensor causal_mask;
-    bool select_nfltmax_at_0;  // set attn_score to -FLT_MAX when causal_mask[...] equal to this
+    bool select_nfltmax_at_0 = false;  // set attn_score to -FLT_MAX when causal_mask[...] equal to this
     void set_causal_mask(PlainTensor mask, bool _select_nfltmax_at_0) {
         causal_mask = mask;
         select_nfltmax_at_0 = _select_nfltmax_at_0;