@fugood/llama.node 1.1.5 → 1.1.6

package/src/llama.cpp/src/llama-model.cpp

@@ -109,8 +109,10 @@ const char * llm_type_name(llm_type type) {
         case LLM_TYPE_A13B:          return "A13B";
         case LLM_TYPE_21B_A3B:       return "21B.A3B";
         case LLM_TYPE_30B_A3B:       return "30B.A3B";
+        case LLM_TYPE_106B_A12B:     return "106B.A12B";
         case LLM_TYPE_235B_A22B:     return "235B.A22B";
         case LLM_TYPE_300B_A47B:     return "300B.A47B";
+        case LLM_TYPE_355B_A32B:     return "355B.A32B";
         case LLM_TYPE_E2B:           return "E2B";
         case LLM_TYPE_E4B:           return "E4B";
         default:                     return "?B";
@@ -190,6 +192,13 @@ static bool weight_buft_supported(const llama_hparams & hparams, ggml_tensor * w
                 ggml_tensor * a = ggml_new_tensor_4d(ctx, GGML_TYPE_F32, w->ne[0], w->ne[1], w->ne[2], w->ne[3]);
                 op_tensor = ggml_add(ctx, a, w);
             } break;
+        case GGML_OP_ADD_ID:
+            {
+                int n_expert_used = hparams.n_expert_used;
+                ggml_tensor * a = ggml_new_tensor_3d(ctx, GGML_TYPE_F32, w->ne[0], n_expert_used, 512);
+                ggml_tensor * c = ggml_new_tensor_2d(ctx, GGML_TYPE_I32, n_expert_used, 512);
+                op_tensor = ggml_add_id(ctx, a, w, c);
+            } break;
         case GGML_OP_MUL:
             {
                 ggml_tensor * a = ggml_new_tensor_4d(ctx, GGML_TYPE_F32, w->ne[0], w->ne[1], w->ne[2], w->ne[3]);
@@ -258,6 +267,10 @@ static bool weight_buft_supported(const llama_hparams & hparams, ggml_tensor * w
                 ggml_tensor * b = ggml_new_tensor_4d(ctx, GGML_TYPE_F32, n_embd, w->ne[1], 1, 1);
                 op_tensor = ggml_im2col(ctx, w, b, 1, 0, 0, 0, 1, 0, false, GGML_TYPE_F16);
             } break;
+        case GGML_OP_SCALE:
+            {
+                op_tensor = ggml_scale(ctx, w, 1.0f);
+            } break;
         default:
             GGML_ABORT("%s: missing test for op %s for tensor %s", __func__, ggml_op_name(op), w->name);
     }
@@ -1434,6 +1447,34 @@ void llama_model::load_hparams(llama_model_loader & ml) {
                     default: type = LLM_TYPE_UNKNOWN;
                 }
             } break;
+        case LLM_ARCH_GLM4_MOE:
+            {
+                ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH,  hparams.n_ff_exp);
+                ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+
+                // MoE parameters
+                ml.get_key(LLM_KV_EXPERT_COUNT,                hparams.n_expert);
+                ml.get_key(LLM_KV_EXPERT_USED_COUNT,           hparams.n_expert_used);
+                ml.get_key(LLM_KV_EXPERT_SHARED_COUNT,         hparams.n_expert_shared);
+                ml.get_key(LLM_KV_LEADING_DENSE_BLOCK_COUNT,   hparams.n_layer_dense_lead, false);
+                ml.get_key(LLM_KV_EXPERT_WEIGHTS_SCALE,        hparams.expert_weights_scale);
+                ml.get_key(LLM_KV_EXPERT_WEIGHTS_NORM,         hparams.expert_weights_norm, false);
+
+                // Expert gating function (GLM-4.5 uses sigmoid)
+                ml.get_key(LLM_KV_EXPERT_GATING_FUNC,          hparams.expert_gating_func, false);
+                if (hparams.expert_gating_func == LLAMA_EXPERT_GATING_FUNC_TYPE_NONE) {
+                    hparams.expert_gating_func = LLAMA_EXPERT_GATING_FUNC_TYPE_SIGMOID;
+                }
+
+                // NextN/MTP parameters
+                ml.get_key(LLM_KV_NEXTN_PREDICT_LAYERS,        hparams.nextn_predict_layers, false);
+
+                switch (hparams.n_layer) {
+                    case 47: type = LLM_TYPE_106B_A12B; break; // GLM-4.5-Air (46 layers + 1 NextN layer)
+                    case 93: type = LLM_TYPE_355B_A32B; break; // GLM-4.5 (92 layers + 1 NextN layer)
+                    default: type = LLM_TYPE_UNKNOWN;
+                }
+            } break;
         case LLM_ARCH_BITNET:
             {
                 ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
@@ -1783,6 +1824,17 @@ void llama_model::load_hparams(llama_model_loader & ml) {
                     default: type = LLM_TYPE_UNKNOWN;
                 }
             } break;
+        case LLM_ARCH_OPENAI_MOE:
+            {
+                ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+                ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH,  hparams.n_ff_exp);
+                ml.get_key(LLM_KV_ATTENTION_SLIDING_WINDOW,    hparams.n_swa);
+
+                hparams.swa_type = LLAMA_SWA_TYPE_STANDARD;
+                hparams.set_swa_pattern(2);
+
+                // TODO: switch (hparams.n_layer)
+            } break;
         case LLM_ARCH_LFM2:
             {
                 ml.get_key(LLM_KV_SHORTCONV_L_CACHE,           hparams.n_shortconv_l_cache);
@@ -1949,6 +2001,7 @@ bool llama_model::load_tensors(llama_model_loader & ml) {
 
     const auto TENSOR_DUPLICATED   = llama_model_loader::TENSOR_DUPLICATED;
     const auto TENSOR_NOT_REQUIRED = llama_model_loader::TENSOR_NOT_REQUIRED;
+    const auto TENSOR_SKIP         = llama_model_loader::TENSOR_SKIP;
 
     // create tensors for the weights
     {
@@ -2004,7 +2057,7 @@ bool llama_model::load_tensors(llama_model_loader & ml) {
             }
 
             // skip unused tensors
-            if (info.op == GGML_OP_NONE) {
+            if (info.op == GGML_OP_NONE || flags & TENSOR_SKIP) {
                 const size_t nbytes = ggml_nbytes(t_meta);
                 LLAMA_LOG_WARN("model has unused tensor %s (size = %zu bytes) -- ignoring\n", tn.str().c_str(), nbytes);
 
@@ -2014,11 +2067,15 @@ bool llama_model::load_tensors(llama_model_loader & ml) {
                 return nullptr;
             }
 
-            // tensors with "bias" suffix are always used with GGML_OP_ADD
+            // tensors with "bias" suffix are always used with GGML_OP_ADD or GGML_OP_ADD_ID
             ggml_op op;
             bool bias = tn.suffix != nullptr && strcmp(tn.suffix, "bias") == 0;
             if (bias) {
-                op = GGML_OP_ADD;
+                if (info.op == GGML_OP_MUL_MAT_ID) {
+                    op = GGML_OP_ADD_ID;
+                } else {
+                    op = GGML_OP_ADD;
+                }
             } else {
                 op = info.op;
             }
@@ -4427,6 +4484,105 @@ bool llama_model::load_tensors(llama_model_loader & ml) {
                         layer.ffn_post_norm  = create_tensor(tn(LLM_TENSOR_FFN_POST_NORM, "weight", i), {n_embd}, 0);
                     }
                 } break;
+            case LLM_ARCH_GLM4_MOE:
+                {
+                    const int64_t n_expert        = hparams.n_expert;
+                    const int64_t n_expert_used   = hparams.n_expert_used;
+                    const int64_t n_expert_shared = hparams.n_expert_shared;
+
+                    GGML_ASSERT(hparams.n_expert > 0 && "n_expert must be > 0 for GLM4_MOE MoE layers");
+                    GGML_ASSERT(hparams.n_expert_used > 0 && "n_expert_used must be > 0 for GLM4_MOE MoE layers");
+
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), { n_embd, n_vocab }, 0);
+
+                    // output
+                    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), { n_embd }, 0);
+                    output = create_tensor(tn(LLM_TENSOR_OUTPUT, "weight"), { n_embd, n_vocab }, TENSOR_NOT_REQUIRED);
+                    // if output is NULL, init from the input tok embed
+                    if (output == NULL) {
+                        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), { n_embd, n_vocab }, TENSOR_DUPLICATED);
+                    }
+
+                    // Load ALL tensors including NextN layer to satisfy total tensor count
+                    // but only PROCESS up to last layer (skipping final NextN layer) in forward pass
+                    for (int i = 0; i < n_layer; ++i) {
+                        int flags = 0;
+                        if (hparams.nextn_predict_layers > 0 && static_cast<uint32_t>(i) >= n_layer - hparams.nextn_predict_layers) {
+                            // skip all tensors in the NextN layers
+                            flags |= TENSOR_SKIP;
+                        }
+
+                        auto & layer = layers[i];
+
+                        layer.attn_norm = create_tensor(tn(LLM_TENSOR_ATTN_NORM, "weight", i), { n_embd }, flags);
+
+                        // GLM-style attention with bias terms
+                        layer.wq = create_tensor(tn(LLM_TENSOR_ATTN_Q, "weight", i), { n_embd, n_embd_head_k * n_head }, flags);
+                        layer.wk = create_tensor(tn(LLM_TENSOR_ATTN_K, "weight", i), { n_embd, n_embd_k_gqa }, flags);
+                        layer.wv = create_tensor(tn(LLM_TENSOR_ATTN_V, "weight", i), { n_embd, n_embd_v_gqa }, flags);
+                        layer.bq = create_tensor(tn(LLM_TENSOR_ATTN_Q, "bias", i), { n_embd_head_k * n_head }, flags);
+                        layer.bk = create_tensor(tn(LLM_TENSOR_ATTN_K, "bias", i), { n_embd_k_gqa }, flags);
+                        layer.bv = create_tensor(tn(LLM_TENSOR_ATTN_V, "bias", i), { n_embd_v_gqa }, flags);
+
+                        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), { n_embd_head_k * n_head, n_embd }, flags);
+
+                        // K/Q norm tensors (optional for GLM-4.5 355B variant)
+                        layer.attn_q_norm = create_tensor(
+                            tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), { n_embd_head_k }, TENSOR_NOT_REQUIRED | flags);
+                        layer.attn_k_norm = create_tensor(
+                            tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), { n_embd_head_k }, TENSOR_NOT_REQUIRED | flags);
+
+                        layer.attn_post_norm = create_tensor(tn(LLM_TENSOR_ATTN_POST_NORM, "weight", i), { n_embd }, flags);
+
+                        // Check if this layer uses MoE or dense FFN based on n_layer_dense_lead
+                        // GLM 4.5 uses hybrid architecture: layer 0 is dense, layers 1+ are MoE
+                        const bool use_moe = (static_cast<uint32_t>(i) >= hparams.n_layer_dense_lead);
+
+                        if (use_moe) {
+                            // MoE layers
+                            layer.ffn_gate_inp =
+                                create_tensor(tn(LLM_TENSOR_FFN_GATE_INP, "weight", i), { n_embd, n_expert }, flags);
+                            layer.ffn_exp_probs_b = create_tensor(tn(LLM_TENSOR_FFN_EXP_PROBS_B, "bias", i), { n_expert }, flags);
+
+                            // MoE branch
+                            const int64_t n_ff_exp = hparams.n_ff_exp ? hparams.n_ff_exp : n_ff / n_expert_used;
+
+                            layer.ffn_gate_exps = create_tensor(
+                                tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), { n_embd, n_ff_exp, n_expert }, flags);
+                            layer.ffn_down_exps = create_tensor(
+                                tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), { n_ff_exp, n_embd, n_expert }, flags);
+                            layer.ffn_up_exps = create_tensor(
+                                tn(LLM_TENSOR_FFN_UP_EXPS, "weight", i), { n_embd, n_ff_exp, n_expert }, flags);
+
+                            // Shared expert
+                            if (n_expert_shared > 0) {
+                                const int64_t n_ff_shexp = n_ff_exp * n_expert_shared;
+                                layer.ffn_gate_shexp = create_tensor(
+                                    tn(LLM_TENSOR_FFN_GATE_SHEXP, "weight", i), { n_embd, n_ff_shexp }, flags);
+                                layer.ffn_down_shexp = create_tensor(
+                                    tn(LLM_TENSOR_FFN_DOWN_SHEXP, "weight", i), { n_ff_shexp, n_embd }, flags);
+                                layer.ffn_up_shexp = create_tensor(
+                                    tn(LLM_TENSOR_FFN_UP_SHEXP, "weight", i), { n_embd, n_ff_shexp }, flags);
+                            }
+                        } else {
+                            // Dense layers (first k layers) - GLM uses separate gate/up projections
+                            layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), { n_embd, n_ff }, flags);
+                            layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), { n_ff, n_embd }, flags);
+                            layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), { n_embd, n_ff }, flags);
+                        }
+
+                        // NextN/MTP tensors (preserved but unused) - conditionally load for last nextn_predict_layers
+                        if (hparams.nextn_predict_layers > 0 && static_cast<uint32_t>(i) >= n_layer - hparams.nextn_predict_layers) {
+                            layer.nextn.eh_proj          = create_tensor(tn(LLM_TENSOR_NEXTN_EH_PROJ, "weight", i), { 2 * n_embd, n_embd }, flags);
+                            layer.nextn.embed_tokens     = create_tensor(tn(LLM_TENSOR_NEXTN_EMBED_TOKENS, "weight", i), { n_embd, n_vocab }, flags);
+                            layer.nextn.enorm            = create_tensor(tn(LLM_TENSOR_NEXTN_ENORM, "weight", i), { n_embd }, flags);
+                            layer.nextn.hnorm            = create_tensor(tn(LLM_TENSOR_NEXTN_HNORM, "weight", i), { n_embd }, flags);
+                            layer.nextn.shared_head_head = create_tensor(tn(LLM_TENSOR_NEXTN_SHARED_HEAD_HEAD, "weight", i), { n_embd, n_vocab }, flags);
+                            layer.nextn.shared_head_norm = create_tensor(tn(LLM_TENSOR_NEXTN_SHARED_HEAD_NORM, "weight", i), { n_embd }, flags);
+                        }
+                    }
+                }
+                break;
             case LLM_ARCH_NEMOTRON:
                 {
                     tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
@@ -5270,6 +5426,46 @@ bool llama_model::load_tensors(llama_model_loader & ml) {
                         layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
                     }
                 } break;
+            case LLM_ARCH_OPENAI_MOE:
+                {
+                    const int64_t n_ff_exp = hparams.n_ff_exp;
+
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+                    // output
+                    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, 0);
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+
+                        layer.attn_norm      = create_tensor(tn(LLM_TENSOR_ATTN_NORM,      "weight", i), {n_embd}, 0);
+                        layer.attn_post_norm = create_tensor(tn(LLM_TENSOR_ATTN_POST_NORM, "weight", i), {n_embd}, 0);
+
+                        layer.wq = create_tensor(tn(LLM_TENSOR_ATTN_Q,   "weight", i), {n_embd, n_head * n_rot}, 0);
+                        layer.wk = create_tensor(tn(LLM_TENSOR_ATTN_K,   "weight", i), {n_embd, n_head_kv * n_rot}, 0);
+                        layer.wv = create_tensor(tn(LLM_TENSOR_ATTN_V,   "weight", i), {n_embd, n_head_kv * n_rot}, 0);
+                        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_head * n_rot, n_embd}, 0);
+
+                        layer.attn_sinks = create_tensor(tn(LLM_TENSOR_ATTN_SINKS, "weight", i), {n_head}, 0);
+
+                        layer.ffn_gate_inp  = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP,  "weight", i), {  n_embd, n_expert}, 0);
+                        layer.ffn_gate_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {  n_embd, n_ff_exp, n_expert}, 0);
+                        layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {n_ff_exp,   n_embd, n_expert}, 0);
+                        layer.ffn_up_exps   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {  n_embd, n_ff_exp, n_expert}, 0);
+
+                        // bias
+                        layer.bq = create_tensor(tn(LLM_TENSOR_ATTN_Q,   "bias", i), {n_head * n_rot}, 0);
+                        layer.bk = create_tensor(tn(LLM_TENSOR_ATTN_K,   "bias", i), {n_head_kv * n_rot}, 0);
+                        layer.bv = create_tensor(tn(LLM_TENSOR_ATTN_V,   "bias", i), {n_head_kv * n_rot}, 0);
+                        layer.bo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd}, 0);
+
+                        layer.ffn_gate_inp_b  = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP,  "bias", i), {n_expert}, 0);
+                        layer.ffn_gate_exps_b = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "bias", i), {n_ff_exp, n_expert}, 0);
+                        layer.ffn_down_exps_b = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "bias", i), {  n_embd, n_expert}, 0);
+                        layer.ffn_up_exps_b   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "bias", i), {n_ff_exp, n_expert}, 0);
+                    }
+                } break;
             case LLM_ARCH_LFM2:
                 {
                     tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
@@ -5642,7 +5838,7 @@ void llama_model::print_info() const {
         LLAMA_LOG_INFO("%s: n_ff_shexp       = %d\n",     __func__, hparams.n_ff_shexp);
     }
 
-    if (arch == LLM_ARCH_QWEN3MOE) {
+    if (arch == LLM_ARCH_QWEN3MOE || arch == LLM_ARCH_OPENAI_MOE) {
         LLAMA_LOG_INFO("%s: n_ff_exp         = %d\n",     __func__, hparams.n_ff_exp);
     }
 
@@ -13564,6 +13760,165 @@ struct llm_build_glm4 : public llm_graph_context {
     }
 };
 
+struct llm_build_glm4_moe : public llm_graph_context {
+    llm_build_glm4_moe(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+        const int64_t n_embd_head = hparams.n_embd_head_v;
+
+        GGML_ASSERT(n_embd_head == hparams.n_embd_head_k);
+
+        ggml_tensor * cur;
+        ggml_tensor * inpL;
+
+        inpL = build_inp_embd(model.tok_embd);
+
+        // inp_pos - contains the positions
+        ggml_tensor * inp_pos = build_inp_pos();
+
+        auto * inp_attn = build_attn_inp_kv_unified();
+
+        ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+        // Only process up to last layer (skip final NextN layer)
+        // Final layer tensors are loaded but not processed in forward pass
+        const int n_transformer_layers = n_layer - hparams.nextn_predict_layers;
+        for (int il = 0; il < n_transformer_layers; ++il) {
+            ggml_tensor * inpSA = inpL;
+
+            // Pre-attention norm
+            cur = build_norm(inpL, model.layers[il].attn_norm, NULL, LLM_NORM_RMS, il);
+            cb(cur, "attn_norm", il);
+
+            // self-attention
+            {
+                ggml_tensor * Qcur = build_lora_mm(model.layers[il].wq, cur);
+                if (model.layers[il].bq) {
+                    Qcur = ggml_add(ctx0, Qcur, model.layers[il].bq);
+                }
+                cb(Qcur, "Qcur", il);
+
+                ggml_tensor * Kcur = build_lora_mm(model.layers[il].wk, cur);
+                if (model.layers[il].bk) {
+                    Kcur = ggml_add(ctx0, Kcur, model.layers[il].bk);
+                }
+                cb(Kcur, "Kcur", il);
+
+                ggml_tensor * Vcur = build_lora_mm(model.layers[il].wv, cur);
+                if (model.layers[il].bv) {
+                    Vcur = ggml_add(ctx0, Vcur, model.layers[il].bv);
+                }
+                cb(Vcur, "Vcur", il);
+
+                Qcur = ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens);
+                Kcur = ggml_reshape_3d(ctx0, Kcur, n_embd_head, n_head_kv, n_tokens);
+                Vcur = ggml_reshape_3d(ctx0, Vcur, n_embd_head, n_head_kv, n_tokens);
+
+                // Apply Q/K norm if available (GLM-4.5 355B variant)
+                if (model.layers[il].attn_q_norm) {
+                    Qcur = build_norm(Qcur, model.layers[il].attn_q_norm, NULL, LLM_NORM_RMS, il);
+                    cb(Qcur, "Qcur_normed", il);
+                }
+                if (model.layers[il].attn_k_norm) {
+                    Kcur = build_norm(Kcur, model.layers[il].attn_k_norm, NULL, LLM_NORM_RMS, il);
+                    cb(Kcur, "Kcur_normed", il);
+                }
+
+                Qcur = ggml_rope_ext(
+                        ctx0, Qcur, inp_pos, nullptr,
+                        n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                        ext_factor, attn_factor, beta_fast, beta_slow
+                        );
+
+                Kcur = ggml_rope_ext(
+                        ctx0, Kcur, inp_pos, nullptr,
+                        n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                        ext_factor, attn_factor, beta_fast, beta_slow
+                        );
+
+                cb(Qcur, "Qcur", il);
+                cb(Kcur, "Kcur", il);
+                cb(Vcur, "Vcur", il);
+
+                cur = build_attn(inp_attn,
+                        model.layers[il].wo, NULL,
+                        Qcur, Kcur, Vcur, nullptr, nullptr, 1.0f/sqrtf(float(n_embd_head)), il);
+            }
+
+            if (il == n_transformer_layers - 1 && inp_out_ids) {
+                cur   = ggml_get_rows(ctx0, cur, inp_out_ids);
+                inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+            }
+
+            ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+            cb(ffn_inp, "ffn_inp", il);
+
+            // Post-attention norm
+            cur = build_norm(ffn_inp, model.layers[il].attn_post_norm, NULL, LLM_NORM_RMS, il);
+            cb(cur, "post_attn_norm", il);
+
+            // Check if this is a dense layer (n_layer_dense_lead=1, so layer 0 is dense)
+            if (static_cast<uint32_t>(il) < hparams.n_layer_dense_lead) {
+                // Dense FFN layer
+                cur = build_ffn(cur,
+                        model.layers[il].ffn_up,   NULL, NULL,
+                        model.layers[il].ffn_gate, NULL, NULL,
+                        model.layers[il].ffn_down, NULL, NULL,
+                        NULL,
+                        LLM_FFN_SILU, LLM_FFN_PAR, il);
+                cb(cur, "ffn_out", il);
+            } else {
+                // Process routed experts using existing MoE infrastructure
+                ggml_tensor * routed_out = build_moe_ffn(cur,
+                        model.layers[il].ffn_gate_inp,
+                        model.layers[il].ffn_up_exps,
+                        model.layers[il].ffn_gate_exps,
+                        model.layers[il].ffn_down_exps,
+                        model.layers[il].ffn_exp_probs_b,
+                        n_expert, n_expert_used,
+                        LLM_FFN_SILU, hparams.expert_weights_norm,
+                        true, hparams.expert_weights_scale,
+                        (llama_expert_gating_func_type) hparams.expert_gating_func,
+                        il);
+                cb(routed_out, "ffn_moe_out", il);
+
+                // Process shared expert on original input
+                ggml_tensor * shared_out = build_ffn(cur,
+                        model.layers[il].ffn_up_shexp,   NULL, NULL,
+                        model.layers[il].ffn_gate_shexp, NULL, NULL,
+                        model.layers[il].ffn_down_shexp, NULL, NULL,
+                        NULL,
+                        LLM_FFN_SILU, LLM_FFN_PAR, il);
+                cb(shared_out, "ffn_shexp_out", il);
+
+                // Final output: routed_output + shared_output
+                cur = ggml_add(ctx0, routed_out, shared_out);
+                cb(cur, "ffn_out", il);
+            }
+
+            cur = ggml_add(ctx0, cur, ffn_inp);
+
+            cur = build_cvec(cur, il);
+            cb(cur, "l_out", il);
+
+            // input for next layer
+            inpL = cur;
+        }
+
+        cur = inpL;
+        cur = build_norm(cur, model.output_norm, NULL, LLM_NORM_RMS, -1);
+
+        cb(cur, "result_norm", -1);
+        res->t_embd = cur;
+
+        // lm_head
+        cur = build_lora_mm(model.output, cur);
+
+        cb(cur, "result_output", -1);
+        res->t_logits = cur;
+
+        ggml_build_forward_expand(gf, cur);
+    }
+};
+
 struct llm_build_nemotron : public llm_graph_context {
     llm_build_nemotron(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
         const int64_t n_embd_head = hparams.n_embd_head_v;
@@ -17251,6 +17606,136 @@ struct llm_build_smollm3 : public llm_graph_context {
     }
 };
 
+struct llm_build_openai_moe_iswa : public llm_graph_context {
+    llm_build_openai_moe_iswa(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+        ggml_tensor * cur;
+        ggml_tensor * inpL;
+
+        inpL = build_inp_embd(model.tok_embd);
+
+        // inp_pos - contains the positions
+        ggml_tensor * inp_pos = build_inp_pos();
+
+        auto * inp_attn = build_attn_inp_kv_unified_iswa();
+
+        for (int il = 0; il < n_layer; ++il) {
+            ggml_tensor * inpSA = inpL;
+
+            // norm
+            cur = build_norm(inpL,
+                    model.layers[il].attn_norm, nullptr,
+                    LLM_NORM_RMS, il);
+            cb(cur, "attn_norm", il);
+
+            // self-attention
+            {
+                // compute Q and K and RoPE them
+                ggml_tensor * Qcur = build_lora_mm(model.layers[il].wq, cur);
+                cb(Qcur, "Qcur", il);
+                if (model.layers[il].bq) {
+                    Qcur = ggml_add(ctx0, Qcur, model.layers[il].bq);
+                    cb(Qcur, "Qcur", il);
+                }
+
+                ggml_tensor * Kcur = build_lora_mm(model.layers[il].wk, cur);
+                cb(Kcur, "Kcur", il);
+                if (model.layers[il].bk) {
+                    Kcur = ggml_add(ctx0, Kcur, model.layers[il].bk);
+                    cb(Kcur, "Kcur", il);
+                }
+
+                ggml_tensor * Vcur = build_lora_mm(model.layers[il].wv, cur);
+                cb(Vcur, "Vcur", il);
+                if (model.layers[il].bv) {
+                    Vcur = ggml_add(ctx0, Vcur, model.layers[il].bv);
+                    cb(Vcur, "Vcur", il);
+                }
+
+                Qcur = ggml_reshape_3d(ctx0, Qcur, n_rot, n_head,    n_tokens);
+                Kcur = ggml_reshape_3d(ctx0, Kcur, n_rot, n_head_kv, n_tokens);
+                Vcur = ggml_reshape_3d(ctx0, Vcur, n_rot, n_head_kv, n_tokens);
+
+                Qcur = ggml_rope_ext(
+                        ctx0, Qcur, inp_pos, nullptr,
+                        n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                        ext_factor, attn_factor, beta_fast, beta_slow
+                        );
+
+                Kcur = ggml_rope_ext(
+                        ctx0, Kcur, inp_pos, nullptr,
+                        n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                        ext_factor, attn_factor, beta_fast, beta_slow
+                        );
+
+                cb(Qcur, "Qcur", il);
+                cb(Kcur, "Kcur", il);
+                cb(Vcur, "Vcur", il);
+
+                cur = build_attn_with_sinks(inp_attn,
+                        model.layers[il].wo, model.layers[il].bo,
+                        Qcur, Kcur, Vcur, nullptr, nullptr, model.layers[il].attn_sinks, 1.0f/sqrtf(float(n_rot)), il);
+
+                cb(cur, "attn_out", il);
+            }
+
+            if (il == n_layer - 1) {
+                // skip computing output for unused tokens
+                ggml_tensor * inp_out_ids = build_inp_out_ids();
+                cur   = ggml_get_rows(ctx0,   cur, inp_out_ids);
+                inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+            }
+
+            ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+            cb(ffn_inp, "ffn_inp", il);
+
+            cur = ffn_inp;
+            cur = build_norm(cur,
+                    model.layers[il].attn_post_norm, nullptr,
+                    LLM_NORM_RMS, il);
+            cb(cur, "attn_post_norm", il);
+
+            // MoE branch
+            cur = build_moe_ffn(cur,
+                    model.layers[il].ffn_gate_inp,  model.layers[il].ffn_gate_inp_b,
+                    model.layers[il].ffn_up_exps,   model.layers[il].ffn_up_exps_b,
+                    model.layers[il].ffn_gate_exps, model.layers[il].ffn_gate_exps_b,
+                    model.layers[il].ffn_down_exps, model.layers[il].ffn_down_exps_b,
+                    nullptr,
+                    n_expert, n_expert_used,
+                    LLM_FFN_SWIGLU_OAI_MOE, false,
+                    false, 0.0,
+                    LLAMA_EXPERT_GATING_FUNC_TYPE_SOFTMAX_WEIGHT,
+                    il);
+            cb(cur, "ffn_moe_out", il);
+
+            cur = ggml_add(ctx0, cur, ffn_inp);
+
+            cur = build_cvec(cur, il);
+            cb(cur, "l_out", il);
+
+            // input for next layer
+            inpL = cur;
+        }
+
+        cur = inpL;
+
+        cur = build_norm(cur,
+                model.output_norm, NULL,
+                LLM_NORM_RMS, -1);
+
+        cb(cur, "result_norm", -1);
+        res->t_embd = cur;
+
+        // lm_head
+        cur = build_lora_mm(model.output, cur);
+
+        cb(cur, "result_output", -1);
+        res->t_logits = cur;
+
+        ggml_build_forward_expand(gf, cur);
+    }
+};
+
 struct llm_build_lfm2 : public llm_graph_context {
     const llama_model & model;
 
@@ -17877,6 +18362,10 @@ ggml_cgraph * llama_model::build_graph(const llm_graph_params & params) const {
             {
                 llm = std::make_unique<llm_build_glm4>(*this, params);
             } break;
+        case LLM_ARCH_GLM4_MOE:
+            {
+                llm = std::make_unique<llm_build_glm4_moe>(*this, params);
+            } break;
         case LLM_ARCH_BITNET:
             {
                 llm = std::make_unique<llm_build_bitnet>(*this, params);
@@ -17990,6 +18479,10 @@ ggml_cgraph * llama_model::build_graph(const llm_graph_params & params) const {
             {
                 llm = std::make_unique<llm_build_smollm3>(*this, params);
             } break;
+        case LLM_ARCH_OPENAI_MOE:
+            {
+                llm = std::make_unique<llm_build_openai_moe_iswa>(*this, params);
+            } break;
         case LLM_ARCH_FALCON_H1:
             {
                 llm = std::make_unique<llm_build_falcon_h1>(*this, params);
@@ -18205,9 +18698,11 @@ llama_rope_type llama_model_rope_type(const llama_model * model) {
         case LLM_ARCH_MINICPM3:
         case LLM_ARCH_DOTS1:
         case LLM_ARCH_HUNYUAN_MOE:
+        case LLM_ARCH_OPENAI_MOE:
         case LLM_ARCH_HUNYUAN_DENSE:
         case LLM_ARCH_LFM2:
         case LLM_ARCH_SMALLTHINKER:
+        case LLM_ARCH_GLM4_MOE:
             return LLAMA_ROPE_TYPE_NEOX;
 
         case LLM_ARCH_QWEN2VL:

package/src/llama.cpp/src/llama-model.h

@@ -101,8 +101,10 @@ enum llm_type {
     LLM_TYPE_A13B,
     LLM_TYPE_21B_A3B, // Ernie MoE small
     LLM_TYPE_30B_A3B,
+    LLM_TYPE_106B_A12B, // GLM-4.5-Air
     LLM_TYPE_235B_A22B,
     LLM_TYPE_300B_A47B, // Ernie MoE big
+    LLM_TYPE_355B_A32B, // GLM-4.5
     LLM_TYPE_E2B,
     LLM_TYPE_E4B,
 };
@@ -166,6 +168,15 @@ struct llama_layer_shortconv {
     struct ggml_tensor * out_proj = nullptr;
 };
 
+struct llama_layer_nextn {
+    struct ggml_tensor * eh_proj          = nullptr;
+    struct ggml_tensor * embed_tokens     = nullptr;
+    struct ggml_tensor * enorm            = nullptr;
+    struct ggml_tensor * hnorm            = nullptr;
+    struct ggml_tensor * shared_head_head = nullptr;
+    struct ggml_tensor * shared_head_norm = nullptr;
+};
+
 struct llama_layer {
     // normalization
     struct ggml_tensor * attn_norm       = nullptr;
@@ -241,10 +252,14 @@ struct llama_layer {
     struct ggml_tensor * ffn_up_enc   = nullptr;
 
     // ff MoE
-    struct ggml_tensor * ffn_gate_inp  = nullptr;
-    struct ggml_tensor * ffn_gate_exps = nullptr;
-    struct ggml_tensor * ffn_down_exps = nullptr;
-    struct ggml_tensor * ffn_up_exps   = nullptr;
+    struct ggml_tensor * ffn_gate_inp    = nullptr;
+    struct ggml_tensor * ffn_gate_exps   = nullptr;
+    struct ggml_tensor * ffn_down_exps   = nullptr;
+    struct ggml_tensor * ffn_up_exps     = nullptr;
+    struct ggml_tensor * ffn_gate_inp_b  = nullptr;
+    struct ggml_tensor * ffn_gate_exps_b = nullptr;
+    struct ggml_tensor * ffn_down_exps_b = nullptr;
+    struct ggml_tensor * ffn_up_exps_b   = nullptr;
 
     // ff shared expert (shexp)
     struct ggml_tensor * ffn_gate_inp_shexp = nullptr;
@@ -349,11 +364,16 @@ struct llama_layer {
     struct ggml_tensor * laurel_r             = nullptr;
     struct ggml_tensor * laurel_post_norm     = nullptr;
 
+    // openai-moe
+    struct ggml_tensor * attn_sinks = nullptr;
+
     struct llama_layer_posnet posnet;
 
     struct llama_layer_convnext convnext;
 
     struct llama_layer_shortconv shortconv;
+
+    struct llama_layer_nextn nextn;
 };
 
 struct llama_model {