diff --git a/veomni/models/transformers/gpt_oss/__init__.py b/veomni/models/transformers/gpt_oss/__init__.py
new file mode 100644
index 00000000..0364f280
--- /dev/null
+++ b/veomni/models/transformers/gpt_oss/__init__.py
@@ -0,0 +1,20 @@
+import transformers
+from transformers import (
+    AutoConfig,
+    AutoModel,
+    AutoModelForCausalLM,
+)
+
+from .configuration_gpt_oss import GptOssConfig
+from .modeling_gpt_oss import (
+    GptOssForCausalLM,
+    GptOssModel,
+)
+
+
+if transformers.__version__ < "4.55.0":
+    AutoConfig.register("gpt_oss", GptOssConfig)
+    AutoModel.register(GptOssConfig, GptOssModel)
+    AutoModelForCausalLM.register(GptOssConfig, GptOssForCausalLM)
+
+__all__ = ["GptOssConfig", "GptOssModel", "GptOssForCausalLM"]
diff --git a/veomni/models/transformers/gpt_oss/configuration_gpt_oss.py b/veomni/models/transformers/gpt_oss/configuration_gpt_oss.py
new file mode 100644
index 00000000..0b7ec76d
--- /dev/null
+++ b/veomni/models/transformers/gpt_oss/configuration_gpt_oss.py
@@ -0,0 +1,119 @@
+# Copyright 2025 The HuggingFace Team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""openai model configuration"""
+
+from transformers.configuration_utils import PretrainedConfig, layer_type_validation
+from transformers.modeling_rope_utils import rope_config_validation
+
+
+class GptOssConfig(PretrainedConfig):
+    r"""
+    This will yield a configuration to that of the BERT
+    [google-bert/bert-base-uncased](https://huggingface.co/google-bert/bert-base-uncased) architecture.
+
+    """
+
+    model_type = "gpt_oss"
+    base_model_pp_plan = {
+        "embed_tokens": (["input_ids"], ["inputs_embeds"]),
+        "layers": (["hidden_states", "attention_mask"], ["hidden_states"]),
+        "norm": (["hidden_states"], ["hidden_states"]),
+    }
+    base_model_tp_plan = {
+        "layers.*.self_attn.q_proj": "colwise",
+        "layers.*.self_attn.k_proj": "colwise",
+        "layers.*.self_attn.v_proj": "colwise",
+        "layers.*.self_attn.o_proj": "rowwise",
+        "layers.*.self_attn.sinks": "local_rowwise",
+        "layers.*.mlp.experts": "gather",
+        "layers.*.mlp.router": "ep_router",
+        "layers.*.mlp.experts.gate_up_proj": "grouped_gemm",
+        "layers.*.mlp.experts.gate_up_proj_bias": "grouped_gemm",
+        "layers.*.mlp.experts.down_proj": "grouped_gemm",
+        "layers.*.mlp.experts.down_proj_bias": "grouped_gemm",
+    }
+
+    def __init__(
+        self,
+        num_hidden_layers: int = 36,
+        num_local_experts: int = 128,
+        vocab_size: int = 201088,
+        hidden_size: int = 2880,
+        intermediate_size: int = 2880,
+        head_dim: int = 64,
+        num_attention_heads: int = 64,
+        num_key_value_heads: int = 8,
+        sliding_window: int = 128,
+        rope_theta: float = 150000.0,
+        tie_word_embeddings=False,
+        hidden_act: str = "silu",
+        initializer_range: float = 0.02,
+        max_position_embeddings=131072,
+        rms_norm_eps: float = 1e-5,
+        rope_scaling={"rope_type": "yarn", "factor": 32.0, "beta_fast": 32.0, "beta_slow": 1.0, "truncate": False},
+        attention_dropout: float = 0.0,
+        num_experts_per_tok=4,
+        router_aux_loss_coef: float = 0.9,
+        output_router_logits=False,
+        use_cache=True,
+        layer_types=None,
+        _moe_implementation="eager",
+        **kwargs,
+    ):
+        self.vocab_size = vocab_size
+        self.hidden_size = hidden_size
+        self.intermediate_size = intermediate_size
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+        self.num_local_experts = num_local_experts
+        self.sliding_window = sliding_window
+        self.num_experts_per_tok = num_experts_per_tok
+        # for backward compatibility
+        if num_key_value_heads is None:
+            num_key_value_heads = num_attention_heads
+
+        self.num_key_value_heads = num_key_value_heads
+        self.hidden_act = hidden_act
+        self.initializer_range = initializer_range
+        self.rms_norm_eps = rms_norm_eps
+        self.rope_theta = rope_theta
+        self.rope_scaling = rope_scaling
+        self.attention_dropout = attention_dropout
+        self.head_dim = head_dim if head_dim is not None else self.hidden_size // self.num_attention_heads
+        self.layer_types = layer_types
+        if self.layer_types is None:
+            self.layer_types = [
+                "sliding_attention" if bool((i + 1) % 2) else "full_attention" for i in range(self.num_hidden_layers)
+            ]
+        layer_type_validation(self.layer_types)
+
+        # Validate the correctness of rotary position embeddings parameters
+        # BC: if there is a 'type' field, copy it it to 'rope_type'.
+        if self.rope_scaling is not None and "type" in self.rope_scaling:
+            self.rope_scaling["rope_type"] = self.rope_scaling["type"]
+        rope_config_validation(self)
+
+        self.attention_bias = True
+        self.max_position_embeddings = max_position_embeddings
+        self.router_aux_loss_coef = router_aux_loss_coef
+        self.output_router_logits = output_router_logits
+        self.use_cache = use_cache
+        self._moe_implementation = _moe_implementation
+        super().__init__(
+            tie_word_embeddings=tie_word_embeddings,
+            **kwargs,
+        )
+
+
+__all__ = ["GptOssConfig"]
diff --git a/veomni/models/transformers/gpt_oss/modeling_gpt_oss.py b/veomni/models/transformers/gpt_oss/modeling_gpt_oss.py
new file mode 100644
index 00000000..30117fe7
--- /dev/null
+++ b/veomni/models/transformers/gpt_oss/modeling_gpt_oss.py
@@ -0,0 +1,721 @@
+#                🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
+#           This file was automatically generated from src/transformers/models/gpt_oss/modular_gpt_oss.py.
+#               Do NOT edit this file manually as any edits will be overwritten by the generation of
+#             the file from the modular. If any change should be done, please apply the change to the
+#                          modular_gpt_oss.py file directly. One of our CI enforces this.
+#                🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
+# coding=utf-8
+# Copyright 2025 The HuggingFace Team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+from typing import Callable, Optional, Union
+
+import torch
+from torch import nn
+from torch.nn import functional as F
+from transformers.cache_utils import Cache, DynamicCache
+from transformers.generation import GenerationMixin
+from transformers.masking_utils import create_causal_mask, create_sliding_window_causal_mask
+from transformers.modeling_outputs import MoeCausalLMOutputWithPast, MoeModelOutputWithPast
+from transformers.modeling_rope_utils import ROPE_INIT_FUNCTIONS, dynamic_rope_update
+from transformers.modeling_utils import ALL_ATTENTION_FUNCTIONS, PreTrainedModel
+from transformers.processing_utils import Unpack
+from transformers.utils import TransformersKwargs, auto_docstring, can_return_tuple
+from transformers.utils.generic import OutputRecorder, check_model_inputs
+
+from ....ops.gpt_fused_moe import fused_moe_forward
+from ...module_utils import GradientCheckpointingLayer
+from .configuration_gpt_oss import GptOssConfig
+
+
+class GptOssRMSNorm(nn.Module):
+    def __init__(self, hidden_size, eps=1e-6):
+        """
+        GptOssRMSNorm is equivalent to T5LayerNorm
+        """
+        super().__init__()
+        self.weight = nn.Parameter(torch.ones(hidden_size))
+        self.variance_epsilon = eps
+
+    def forward(self, hidden_states):
+        input_dtype = hidden_states.dtype
+        hidden_states = hidden_states.to(torch.float32)
+        variance = hidden_states.pow(2).mean(-1, keepdim=True)
+        hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)
+        return (self.weight * hidden_states).to(input_dtype)  # main diff with Llama
+
+    def extra_repr(self):
+        return f"{tuple(self.weight.shape)}, eps={self.variance_epsilon}"
+
+
+class GptOssExperts(nn.Module):
+    def __init__(self, config: GptOssConfig):
+        super().__init__()
+        self.intermediate_size = config.intermediate_size
+        self.num_experts = config.num_local_experts
+        self.top_k = config.num_experts_per_tok
+        self.hidden_size = config.hidden_size
+        self.moe_implementation = config._moe_implementation
+        self.expert_dim = self.intermediate_size
+        self.gate_up_proj = nn.Parameter(torch.empty(self.num_experts, self.hidden_size, 2 * self.expert_dim))
+        self.gate_up_proj_bias = nn.Parameter(torch.empty(self.num_experts, 2 * self.expert_dim))
+        self.down_proj = nn.Parameter(torch.empty((self.num_experts, self.expert_dim, self.hidden_size)))
+        self.down_proj_bias = nn.Parameter(torch.empty(self.num_experts, self.hidden_size))
+        self.alpha = 1.702
+        self.limit = 7.0
+
+    def forward(self, hidden_states: torch.Tensor, router_indices=None, routing_weights=None) -> torch.Tensor:
+        """
+        When training is is more efficient to just loop over the experts and compute the output for each expert
+        as otherwise the memory would explode.
+
+        For inference we can sacrifice some memory and compute the output for all experts at once. By repeating the inputs.
+
+        Args:
+            hidden_states (torch.Tensor): (batch_size, seq_len, hidden_size)
+            selected_experts (torch.Tensor): (batch_size * token_num, top_k)
+            routing_weights (torch.Tensor): (batch_size * token_num, top_k)
+        Returns:
+            torch.Tensor
+        """
+        batch_size = hidden_states.shape[0]
+        hidden_states = hidden_states.reshape(-1, self.hidden_size)  # (num_tokens, hidden_size)
+        num_experts = self.num_experts
+        if self.training:
+            if self.moe_implementation == "eager":
+                routing_weights = hidden_states.new_zeros((hidden_states.shape[0], num_experts)).scatter_(1, router_indices, routing_weights)
+                next_states = torch.zeros_like(hidden_states, dtype=hidden_states.dtype, device=hidden_states.device)
+                with torch.no_grad():
+                    expert_mask = torch.nn.functional.one_hot(router_indices, num_classes=num_experts)
+                    expert_mask = expert_mask.permute(2, 1, 0)
+                    # we sum on the top_k and on the sequence lenght to get which experts
+                    # are hit this time around
+                    expert_hitted = torch.greater(expert_mask.sum(dim=(-1, -2)), 0).nonzero()
+                for expert_idx in expert_hitted[:]:
+                    with torch.no_grad():
+                        _, token_idx = torch.where(expert_mask[expert_idx[0]])
+                    current_state = hidden_states[token_idx]
+                    gate_up = current_state @ self.gate_up_proj[expert_idx] + self.gate_up_proj_bias[expert_idx]
+                    gate, up = gate_up[..., ::2], gate_up[..., 1::2]
+                    gate = gate.clamp(min=None, max=self.limit)
+                    up = up.clamp(min=-self.limit, max=self.limit)
+                    glu = gate * torch.sigmoid(gate * self.alpha)
+                    gated_output = (up + 1) * glu
+                    out = gated_output @ self.down_proj[expert_idx] + self.down_proj_bias[expert_idx]
+                    weighted_output = out[0] * routing_weights[token_idx, expert_idx, None]
+                    next_states.index_add_(0, token_idx, weighted_output.to(hidden_states.dtype))
+                next_states = next_states.view(batch_size, -1, self.hidden_size)
+
+            elif self.moe_implementation == "fused":
+                next_states = fused_moe_forward(
+                    module=self,
+                    num_experts=self.num_experts,
+                    routing_weights=routing_weights,
+                    selected_experts=router_indices,
+                    hidden_states=hidden_states,
+                    gate_up_proj=self.gate_up_proj,
+                    gate_up_proj_bias=self.gate_up_proj_bias,
+                    down_proj=self.down_proj,
+                    down_proj_bias=self.down_proj_bias,
+                    limit=self.limit,
+                    alpha=self.alpha,
+                ).view(batch_size, -1, self.hidden_size)
+            else:
+                raise NotImplementedError
+        else:
+            hidden_states = hidden_states.repeat(num_experts, 1)
+            hidden_states = hidden_states.view(num_experts, -1, self.hidden_size)
+            gate_up = torch.bmm(hidden_states, self.gate_up_proj) + self.gate_up_proj_bias[..., None, :]
+            gate, up = gate_up[..., ::2], gate_up[..., 1::2]
+            gate = gate.clamp(min=None, max=self.limit)
+            up = up.clamp(min=-self.limit, max=self.limit)
+            glu = gate * torch.sigmoid(gate * self.alpha)
+            next_states = torch.bmm(((up + 1) * glu), self.down_proj)
+            next_states = next_states + self.down_proj_bias[..., None, :]
+            next_states = next_states.view(num_experts, batch_size, -1, self.hidden_size)
+            next_states = next_states * routing_weights.transpose(0, 1).view(num_experts, batch_size, -1)[..., None]
+            next_states = next_states.sum(dim=0)
+        return next_states
+
+
+class GptOssTopKRouter(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.top_k = config.num_experts_per_tok
+        self.num_experts = config.num_local_experts
+        self.hidden_dim = config.hidden_size
+        self.weight = nn.Parameter(torch.empty(self.num_experts, self.hidden_dim))
+        self.bias = nn.Parameter(torch.empty(self.num_experts))
+
+    def forward(self, hidden_states):
+        hidden_states = hidden_states.reshape(-1, self.hidden_dim)
+        router_logits = F.linear(hidden_states, self.weight, self.bias)  # (seq_len, num_experts)
+        router_top_value, router_indices = torch.topk(router_logits, self.top_k, dim=-1)  # (seq_len, top_k)
+        router_top_value = torch.nn.functional.softmax(router_top_value, dim=1, dtype=router_top_value.dtype)
+        # router_scores = torch.zeros_like(router_logits).scatter_(1, router_indices, router_top_value)
+        return router_top_value, router_indices
+
+
+class GptOssMLP(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.router = GptOssTopKRouter(config)
+        self.experts = GptOssExperts(config)
+
+    def forward(self, hidden_states):
+        router_top_value, router_indices = self.router(hidden_states)  # (num_experts, seq_len)
+        routed_out = self.experts(hidden_states, router_indices=router_indices, routing_weights=router_top_value)
+        return routed_out, router_top_value
+
+
+class GptOssRotaryEmbedding(nn.Module):
+    def __init__(self, config: GptOssConfig, device=None):
+        super().__init__()
+        # BC: "rope_type" was originally "type"
+        if hasattr(config, "rope_scaling") and isinstance(config.rope_scaling, dict):
+            self.rope_type = config.rope_scaling.get("rope_type", config.rope_scaling.get("type"))
+        else:
+            self.rope_type = "default"
+        self.max_seq_len_cached = config.max_position_embeddings
+        self.original_max_seq_len = config.max_position_embeddings
+
+        self.config = config
+        self.rope_init_fn = ROPE_INIT_FUNCTIONS[self.rope_type]
+
+        inv_freq, self.attention_scaling = self.rope_init_fn(self.config, device)
+        self.register_buffer("inv_freq", inv_freq, persistent=False)
+        self.original_inv_freq = self.inv_freq
+
+    @torch.no_grad()
+    @dynamic_rope_update  # power user: used with advanced RoPE types (e.g. dynamic rope)
+    def forward(self, x, position_ids):
+        inv_freq_expanded = self.inv_freq[None, :, None].float().expand(position_ids.shape[0], -1, 1).to(x.device)
+        position_ids_expanded = position_ids[:, None, :].float()
+
+        device_type = x.device.type if isinstance(x.device.type, str) and x.device.type != "mps" else "cpu"
+        with torch.autocast(device_type=device_type, enabled=False):  # Force float32
+            freqs = (inv_freq_expanded.float() @ position_ids_expanded.float()).transpose(1, 2)
+            emb = freqs
+            cos = emb.cos() * self.attention_scaling
+            sin = emb.sin() * self.attention_scaling
+
+        return cos.to(x.dtype), sin.to(x.dtype)
+
+
+def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
+    """
+    This is the equivalent of torch.repeat_interleave(x, dim=1, repeats=n_rep). The hidden states go from (batch,
+    num_key_value_heads, seqlen, head_dim) to (batch, num_attention_heads, seqlen, head_dim)
+    """
+    batch, num_key_value_heads, slen, head_dim = hidden_states.shape
+    if n_rep == 1:
+        return hidden_states
+    hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)
+    return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)
+
+
+def _apply_rotary_emb(
+    x: torch.Tensor,
+    cos: torch.Tensor,
+    sin: torch.Tensor,
+) -> torch.Tensor:
+    first_half, second_half = torch.chunk(x, 2, dim=-1)
+    first_ = first_half * cos - second_half * sin
+    second_ = second_half * cos + first_half * sin
+    return torch.cat((first_, second_), dim=-1)
+
+
+def apply_rotary_pos_emb(q, k, cos, sin, position_ids=None, unsqueeze_dim=1):
+    cos = cos.unsqueeze(unsqueeze_dim)
+    sin = sin.unsqueeze(unsqueeze_dim)
+    q_embed = _apply_rotary_emb(q, cos, sin)
+    k_embed = _apply_rotary_emb(k, cos, sin)
+    return q_embed, k_embed
+
+
+def eager_attention_forward(
+    module: nn.Module,
+    query: torch.Tensor,
+    key: torch.Tensor,
+    value: torch.Tensor,
+    attention_mask: Optional[torch.Tensor],
+    scaling: float,
+    dropout: float = 0.0,
+    **kwargs,
+):
+    key_states = repeat_kv(key, module.num_key_value_groups)
+    value_states = repeat_kv(value, module.num_key_value_groups)
+    attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling
+    if attention_mask is not None:
+        causal_mask = attention_mask[:, :, :, : key_states.shape[-2]]
+        attn_weights = attn_weights + causal_mask
+
+    sinks = module.sinks.reshape(1, -1, 1, 1).expand(query.shape[0], -1, query.shape[-2], -1)
+    combined_logits = torch.cat([attn_weights, sinks], dim=-1)
+
+    # This was not in the original implementation and slightly affect results; it prevents overflow in BF16/FP16
+    # when training with bsz>1 we clamp max values.
+
+    combined_logits = combined_logits - combined_logits.max(dim=-1, keepdim=True).values
+    probs = F.softmax(combined_logits, dim=-1, dtype=combined_logits.dtype)
+    scores = probs[..., :-1]  # we drop the sink here
+    attn_weights = nn.functional.dropout(scores, p=dropout, training=module.training)
+    attn_output = torch.matmul(attn_weights, value_states)
+    attn_output = attn_output.transpose(1, 2).contiguous()
+    return attn_output, attn_weights
+
+
+class GptOssAttention(nn.Module):
+    """Multi-headed attention from 'Attention Is All You Need' paper"""
+
+    def __init__(self, config: GptOssConfig, layer_idx: int):
+        super().__init__()
+        self.config = config
+        self.layer_idx = layer_idx
+        self.head_dim = getattr(config, "head_dim", config.hidden_size // config.num_attention_heads)
+        self.num_key_value_groups = config.num_attention_heads // config.num_key_value_heads
+        self.scaling = self.head_dim**-0.5
+        self.attention_dropout = config.attention_dropout
+        self.is_causal = True
+        self.q_proj = nn.Linear(
+            config.hidden_size, config.num_attention_heads * self.head_dim, bias=config.attention_bias
+        )
+        self.k_proj = nn.Linear(
+            config.hidden_size, config.num_key_value_heads * self.head_dim, bias=config.attention_bias
+        )
+        self.v_proj = nn.Linear(
+            config.hidden_size, config.num_key_value_heads * self.head_dim, bias=config.attention_bias
+        )
+        self.o_proj = nn.Linear(
+            config.num_attention_heads * self.head_dim, config.hidden_size, bias=config.attention_bias
+        )
+        self.sliding_window = config.sliding_window if config.layer_types[layer_idx] == "sliding_attention" else None
+        self.sinks = nn.Parameter(torch.empty(config.num_attention_heads))
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        position_embeddings: tuple[torch.Tensor, torch.Tensor],
+        attention_mask: Optional[torch.Tensor],
+        past_key_value: Optional[Cache] = None,
+        cache_position: Optional[torch.LongTensor] = None,
+        **kwargs: Unpack[TransformersKwargs],
+    ) -> tuple[torch.Tensor, torch.Tensor]:
+        input_shape = hidden_states.shape[:-1]
+        hidden_shape = (*input_shape, -1, self.head_dim)
+
+        query_states = self.q_proj(hidden_states).view(hidden_shape).transpose(1, 2)
+        key_states = self.k_proj(hidden_states).view(hidden_shape).transpose(1, 2)
+        value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)
+
+        cos, sin = position_embeddings
+        query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)
+
+        if past_key_value is not None:
+            cache_kwargs = {"cache_position": cache_position}
+            key_states, value_states = past_key_value.update(key_states, value_states, self.layer_idx, cache_kwargs)
+
+        attention_interface: Callable = eager_attention_forward
+        if self.config._attn_implementation != "eager":
+            attention_interface = ALL_ATTENTION_FUNCTIONS[self.config._attn_implementation]
+
+        attn_output, attn_weights = attention_interface(
+            self,
+            query_states,
+            key_states,
+            value_states,
+            attention_mask,
+            dropout=0.0 if not self.training else self.attention_dropout,
+            scaling=self.scaling,
+            sliding_window=self.sliding_window,
+            s_aux=self.sinks,  # diff with Llama
+            **kwargs,
+        )
+
+        attn_output = attn_output.reshape(*input_shape, -1).contiguous()
+        attn_output = self.o_proj(attn_output)
+        return attn_output, attn_weights
+
+
+class GptOssDecoderLayer(GradientCheckpointingLayer):
+    def __init__(self, config: GptOssConfig, layer_idx: int):
+        super().__init__()
+        self.hidden_size = config.hidden_size
+        self.self_attn = GptOssAttention(config=config, layer_idx=layer_idx)
+        self.mlp = GptOssMLP(config)
+        self.input_layernorm = GptOssRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+        self.post_attention_layernorm = GptOssRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+        self.attention_type = config.layer_types[layer_idx]
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_value: Optional[Cache] = None,
+        use_cache: Optional[bool] = False,
+        cache_position: Optional[torch.LongTensor] = None,
+        position_embeddings: Optional[tuple[torch.Tensor, torch.Tensor]] = None,  # necessary, but kept here for BC
+        **kwargs: Unpack[TransformersKwargs],
+    ) -> tuple[torch.Tensor]:
+        residual = hidden_states
+        hidden_states = self.input_layernorm(hidden_states)
+        # Self Attention
+        hidden_states, _ = self.self_attn(
+            hidden_states=hidden_states,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_value=past_key_value,
+            use_cache=use_cache,
+            cache_position=cache_position,
+            position_embeddings=position_embeddings,
+            **kwargs,
+        )
+        hidden_states = residual + hidden_states
+
+        # Fully Connected
+        residual = hidden_states
+        hidden_states = self.post_attention_layernorm(hidden_states)
+        hidden_states, _ = self.mlp(hidden_states)  # diff with llama: router scores
+        hidden_states = residual + hidden_states
+        return hidden_states
+
+
+@auto_docstring
+class GptOssPreTrainedModel(PreTrainedModel):
+    config: GptOssConfig
+    base_model_prefix = "model"
+    supports_gradient_checkpointing = True
+    _no_split_modules = ["GptOssDecoderLayer"]
+    _skip_keys_device_placement = ["past_key_values"]
+    _supports_flash_attn = True
+    _supports_sdpa = False
+    _supports_flex_attn = True
+
+    _can_compile_fullgraph = True
+    _supports_attention_backend = True
+    _can_record_outputs = {
+        "router_logits": OutputRecorder(GptOssTopKRouter, index=0),
+        "hidden_states": GptOssDecoderLayer,
+        "attentions": GptOssAttention,
+    }
+    _keep_in_fp32_modules = ["post_attention_layernorm", "input_layernorm", "norm"]
+    _supports_flash_attention = False
+    _supports_flex_attention = False
+
+    def _init_weights(self, module):
+        std = self.config.initializer_range
+        if isinstance(module, nn.Linear):
+            module.weight.data.normal_(mean=0.0, std=std)
+            if module.bias is not None:
+                module.bias.data.zero_()
+        elif isinstance(module, nn.Parameter):
+            module.data.normal_(mean=0.0, std=std)
+        elif isinstance(module, nn.Embedding):
+            module.weight.data.normal_(mean=0.0, std=std)
+            if module.padding_idx is not None:
+                module.weight.data[module.padding_idx].zero_()
+        elif isinstance(module, GptOssRMSNorm):
+            module.weight.data.fill_(1.0)
+        elif isinstance(module, GptOssExperts):
+            module.gate_up_proj.data.normal_(mean=0.0, std=std)
+            module.gate_up_proj_bias.data.zero_()
+            module.down_proj.data.normal_(mean=0.0, std=std)
+            module.down_proj_bias.data.zero_()
+        elif isinstance(module, GptOssAttention):
+            module.sinks.data.normal_(mean=0.0, std=std)
+        elif isinstance(module, GptOssTopKRouter):
+            module.weight.data.normal_(mean=0.0, std=std)
+            module.bias.data.normal_(mean=0.0, std=std)
+
+
+@auto_docstring
+class GptOssModel(GptOssPreTrainedModel):
+    _no_split_modules = ["GptOssDecoderLayer"]
+
+    def __init__(self, config: GptOssConfig):
+        super().__init__(config)
+        self.padding_idx = config.pad_token_id
+        self.vocab_size = config.vocab_size
+
+        self.embed_tokens = nn.Embedding(config.vocab_size, config.hidden_size, self.padding_idx)
+        self.layers = nn.ModuleList(
+            [GptOssDecoderLayer(config, layer_idx) for layer_idx in range(config.num_hidden_layers)]
+        )
+        self.norm = GptOssRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+        self.rotary_emb = GptOssRotaryEmbedding(config=config)
+        self.gradient_checkpointing = False
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    @check_model_inputs
+    @auto_docstring
+    def forward(
+        self,
+        input_ids: Optional[torch.LongTensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[list[torch.FloatTensor]] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        use_cache: Optional[bool] = None,
+        cache_position: Optional[torch.LongTensor] = None,
+        **kwargs: Unpack[TransformersKwargs],
+    ) -> MoeModelOutputWithPast:
+        if (input_ids is None) ^ (inputs_embeds is not None):
+            raise ValueError("You must specify exactly one of input_ids or inputs_embeds")
+
+        if use_cache and past_key_values is None:
+            past_key_values = DynamicCache()
+
+        if inputs_embeds is None:
+            inputs_embeds = self.embed_tokens(input_ids)
+
+        if cache_position is None:
+            past_seen_tokens = past_key_values.get_seq_length() if past_key_values is not None else 0
+            cache_position = torch.arange(
+                past_seen_tokens, past_seen_tokens + inputs_embeds.shape[1], device=inputs_embeds.device
+            )
+        if position_ids is None:
+            position_ids = cache_position.unsqueeze(0)
+
+        # It may already have been prepared by e.g. `generate`
+        if not isinstance(causal_mask_mapping := attention_mask, dict):
+            mask_kwargs = {
+                "config": self.config,
+                "input_embeds": inputs_embeds,
+                "attention_mask": attention_mask,
+                "cache_position": cache_position,
+                "past_key_values": past_key_values,
+            }
+            causal_mask_mapping = {
+                "full_attention": create_causal_mask(**mask_kwargs),
+                "sliding_attention": create_sliding_window_causal_mask(**mask_kwargs),
+            }
+
+        hidden_states = inputs_embeds
+        position_embeddings = self.rotary_emb(hidden_states, position_ids)
+
+        for decoder_layer in self.layers:
+            hidden_states = decoder_layer(
+                hidden_states,
+                attention_mask=causal_mask_mapping[decoder_layer.attention_type],
+                position_ids=position_ids,
+                past_key_value=past_key_values,
+                use_cache=use_cache,
+                cache_position=cache_position,
+                position_embeddings=position_embeddings,
+                **kwargs,
+            )
+        hidden_states = self.norm(hidden_states)
+        return MoeModelOutputWithPast(
+            last_hidden_state=hidden_states,
+            past_key_values=past_key_values,
+        )
+
+
+def load_balancing_loss_func(
+    gate_logits: Union[torch.Tensor, tuple[torch.Tensor], None],
+    num_experts: Optional[int] = None,
+    top_k=2,
+    attention_mask: Optional[torch.Tensor] = None,
+) -> Union[torch.Tensor, int]:
+    r"""
+    Computes auxiliary load balancing loss as in Switch Transformer - implemented in Pytorch.
+
+    See Switch Transformer (https://huggingface.co/papers/2101.03961) for more details. This function implements the loss
+    function presented in equations (4) - (6) of the paper. It aims at penalizing cases where the routing between
+    experts is too unbalanced.
+
+    Args:
+        gate_logits:
+            Logits from the `gate`, should be a tuple of model.config.num_hidden_layers tensors of
+            shape [batch_size X sequence_length, num_experts].
+        num_experts:
+            Number of experts
+        top_k:
+            The number of experts to route per-token, can be also interpreted as the `top-k` routing
+            parameter.
+        attention_mask (`torch.Tensor`, *optional*):
+            The attention_mask used in forward function
+            shape [batch_size X sequence_length] if not None.
+
+    Returns:
+        The auxiliary loss.
+    """
+    if gate_logits is None or not isinstance(gate_logits, tuple):
+        return 0
+
+    if isinstance(gate_logits, tuple):
+        compute_device = gate_logits[0].device
+        concatenated_gate_logits = torch.cat([layer_gate.to(compute_device) for layer_gate in gate_logits], dim=0)
+
+    routing_weights = torch.nn.functional.softmax(concatenated_gate_logits, dim=-1)
+
+    _, selected_experts = torch.topk(routing_weights, top_k, dim=-1)
+
+    expert_mask = torch.nn.functional.one_hot(selected_experts, num_experts)
+
+    if attention_mask is None:
+        # Compute the percentage of tokens routed to each experts
+        tokens_per_expert = torch.mean(expert_mask.float(), dim=0)
+
+        # Compute the average probability of routing to these experts
+        router_prob_per_expert = torch.mean(routing_weights, dim=0)
+    else:
+        batch_size, sequence_length = attention_mask.shape
+        num_hidden_layers = concatenated_gate_logits.shape[0] // (batch_size * sequence_length)
+
+        # Compute the mask that masks all padding tokens as 0 with the same shape of expert_mask
+        expert_attention_mask = (
+            attention_mask[None, :, :, None, None]
+            .expand((num_hidden_layers, batch_size, sequence_length, top_k, num_experts))
+            .reshape(-1, top_k, num_experts)
+            .to(compute_device)
+        )
+
+        # Compute the percentage of tokens routed to each experts
+        tokens_per_expert = torch.sum(expert_mask.float() * expert_attention_mask, dim=0) / torch.sum(
+            expert_attention_mask, dim=0
+        )
+
+        # Compute the mask that masks all padding tokens as 0 with the same shape of tokens_per_expert
+        router_per_expert_attention_mask = (
+            attention_mask[None, :, :, None]
+            .expand((num_hidden_layers, batch_size, sequence_length, num_experts))
+            .reshape(-1, num_experts)
+            .to(compute_device)
+        )
+
+        # Compute the average probability of routing to these experts
+        router_prob_per_expert = torch.sum(routing_weights * router_per_expert_attention_mask, dim=0) / torch.sum(
+            router_per_expert_attention_mask, dim=0
+        )
+
+    overall_loss = torch.sum(tokens_per_expert * router_prob_per_expert.unsqueeze(0))
+    return overall_loss * num_experts
+
+
+@auto_docstring
+class GptOssForCausalLM(GptOssPreTrainedModel, GenerationMixin):
+    _tied_weights_keys = ["lm_head.weight"]
+    _tp_plan = {"lm_head": "colwise_rep"}
+    _pp_plan = {"lm_head": (["hidden_states"], ["logits"])}
+
+    def __init__(self, config):
+        super().__init__(config)
+        self.model = GptOssModel(config)
+        self.vocab_size = config.vocab_size
+        self.lm_head = nn.Linear(config.hidden_size, config.vocab_size, bias=False)
+        self.router_aux_loss_coef = config.router_aux_loss_coef
+        self.num_experts = config.num_local_experts
+        self.num_experts_per_tok = config.num_experts_per_tok
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def set_decoder(self, decoder):
+        self.model = decoder
+
+    def get_decoder(self):
+        return self.model
+
+    @can_return_tuple
+    @auto_docstring
+    def forward(
+        self,
+        input_ids: Optional[torch.LongTensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[Cache] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        labels: Optional[torch.LongTensor] = None,
+        use_cache: Optional[bool] = None,
+        output_router_logits: Optional[bool] = None,
+        cache_position: Optional[torch.LongTensor] = None,
+        logits_to_keep: Union[int, torch.Tensor] = 0,
+        **kwargs: Unpack[TransformersKwargs],
+    ) -> MoeCausalLMOutputWithPast:
+        r"""
+        labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
+            Labels for computing the masked language modeling loss. Indices should either be in `[0, ...,
+            config.vocab_size]` or -100 (see `input_ids` docstring). Tokens with indices set to `-100` are ignored
+            (masked), the loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`.
+
+        Example:
+
+        ```python
+        >>> from transformers import AutoTokenizer, GptOssForCausalLM
+
+        >>> model = GptOssForCausalLM.from_pretrained("mistralai/GptOss-8x7B-v0.1")
+        >>> tokenizer = AutoTokenizer.from_pretrained("mistralai/GptOss-8x7B-v0.1")
+
+        >>> prompt = "Hey, are you conscious? Can you talk to me?"
+        >>> inputs = tokenizer(prompt, return_tensors="pt")
+
+        >>> # Generate
+        >>> generate_ids = model.generate(inputs.input_ids, max_length=30)
+        >>> tokenizer.batch_decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
+        "Hey, are you conscious? Can you talk to me?\nI'm not conscious, but I can talk to you."
+        ```"""
+
+        output_router_logits = (
+            output_router_logits if output_router_logits is not None else self.config.output_router_logits
+        )
+
+        # decoder outputs consists of (dec_features, layer_state, dec_hidden, dec_attn)
+        outputs: MoeModelOutputWithPast = self.model(
+            input_ids=input_ids,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_values=past_key_values,
+            inputs_embeds=inputs_embeds,
+            use_cache=use_cache,
+            output_router_logits=output_router_logits,
+            cache_position=cache_position,
+            **kwargs,
+        )
+
+        hidden_states = outputs.last_hidden_state
+        # Only compute necessary logits, and do not upcast them to float if we are not computing the loss
+        slice_indices = slice(-logits_to_keep, None) if isinstance(logits_to_keep, int) else logits_to_keep
+        logits = self.lm_head(hidden_states[:, slice_indices, :])
+
+        loss = None
+        if labels is not None:
+            loss = self.loss_function(logits, labels, self.vocab_size, **kwargs)
+
+        aux_loss = None
+        if output_router_logits:
+            aux_loss = load_balancing_loss_func(
+                outputs.router_logits,
+                self.num_experts,
+                self.num_experts_per_tok,
+                attention_mask,
+            )
+            if labels is not None:
+                loss += self.router_aux_loss_coef * aux_loss.to(loss.device)  # make sure to reside in the same device
+
+        return MoeCausalLMOutputWithPast(
+            loss=loss,
+            aux_loss=aux_loss,
+            logits=logits,
+            past_key_values=outputs.past_key_values,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+            router_logits=outputs.router_logits,
+        )
+
+ModelClass = GptOssForCausalLM
+
+__all__ = ["GptOssForCausalLM", "GptOssModel", "GptOssPreTrainedModel"]
diff --git a/veomni/models/transformers/gpt_oss/parallel_plan.py b/veomni/models/transformers/gpt_oss/parallel_plan.py
new file mode 100644
index 00000000..d8d4bda2
--- /dev/null
+++ b/veomni/models/transformers/gpt_oss/parallel_plan.py
@@ -0,0 +1,16 @@
+from torch.distributed._tensor import Shard
+
+from ....distributed.parallel_plan import ParallelPlan
+
+
+def get_paralle_plan():
+    ep_plan = {
+        "model.layers.*.mlp.experts.gate_up_proj_blocks": Shard(0),
+        "model.layers.*.mlp.experts.down_proj_blocks": Shard(0),
+        "model.layers.*.mlp.experts.gate_up_proj_bias": Shard(0),
+        "model.layers.*.mlp.experts.down_proj_bias": Shard(0),
+    }
+    parallel_plan = ParallelPlan(
+        ep_plan=ep_plan,
+    )
+    return parallel_plan
diff --git a/veomni/ops/gpt_fused_moe.py b/veomni/ops/gpt_fused_moe.py
new file mode 100644
index 00000000..29233865
--- /dev/null
+++ b/veomni/ops/gpt_fused_moe.py
@@ -0,0 +1,267 @@
+import torch
+
+from ..utils.import_utils import is_fused_moe_available
+
+
+if is_fused_moe_available():
+    from .group_gemm.kernel.group_gemm import group_gemm_same_mn, group_gemm_same_nk
+    from .group_gemm.kernel.moe import expert_histogram, moe_gather, moe_scatter
+
+
+class FusedMoeExpertFunction(torch.autograd.Function):
+    @staticmethod
+    def forward(
+        ctx,
+        num_experts: int,
+        gate_weights: torch.Tensor,
+        expert_index: torch.Tensor,
+        hidden_states: torch.Tensor,
+        gate_up_proj: torch.Tensor,
+        gate_up_proj_bias: torch.Tensor,
+        down_proj: torch.Tensor,
+        down_proj_bias: torch.Tensor,
+        alpha: float,
+        limit: float,
+    ):
+        # MOE Step 3: 分发输入 tokens 到专家（与原逻辑一致）
+        splits = expert_histogram(expert_index, num_experts)
+        scatter_index = expert_index.flatten().argsort(stable=True).argsort().int().view(expert_index.shape)
+        scatter_output = moe_scatter(hidden_states, scatter_index)  # (total_tokens, hidden_size)
+        cumsum_t = torch.cumsum(splits, dim=0)
+        max_M = scatter_output.shape[0]
+
+        # MOE Step 4: 合并计算 Gate 和 Up 投影（新增逻辑）
+        # 输入: scatter_output (total_tokens, hidden_size)
+        # 权重: gate_up_proj (num_experts, hidden_size, 2*expert_dim)
+        # 偏置: gate_up_proj_bias (num_experts, 2*expert_dim)
+        # 输出: gate_up (total_tokens, 2*expert_dim)
+        gate_up = group_gemm_same_nk(
+            a=scatter_output,
+            b=gate_up_proj,
+            cumsum_M=cumsum_t,
+            max_M=max_M,
+            transpose_a=False,
+            transpose_b=False,
+        )
+        gate_up_bias = gate_up_proj_bias.index_select(0, expert_index.flatten()).view(gate_up.shape)
+        scattered_gate_up_bias = torch.empty_like(gate_up_bias)
+        scattered_gate_up_bias[scatter_index.flatten()] = gate_up_bias
+        gate_up += scattered_gate_up_bias
+
+        # 拆分 Gate 和 Up 分量（新增逻辑）
+        gate = gate_up[..., ::2]  # 取偶数列作为 gate
+        up = gate_up[..., 1::2]   # 取奇数列作为 up
+
+        # MOE Step 5: 应用修正的 SwiGLU 激活（新增逻辑）
+        up = up.clamp(min=-limit, max=limit)            # up 双向截断
+        gate = gate.clamp(min=None, max=limit)            # gate 单向截断
+        gate_activation = gate * torch.sigmoid(gate * alpha)        # 带 alpha 缩放的 sigmoid
+        gate_up_activation = (up + 1) * gate_activation                   # (up + 1) 修正项
+
+        # MOE Step 8: compute the the weighted linear layer 1 result
+        # MOE Step 8-1: compute scattered_gate_weight, shape is (batch_size * sequence_len * topk)
+        reshaped_gate_weight = gate_weights.reshape(-1, 1)
+        scattered_gate_weight = torch.empty_like(reshaped_gate_weight)
+        scattered_gate_weight[scatter_index.flatten()] = reshaped_gate_weight
+
+        # MOE Step 8-2: multiply activate with scattered_gate_weight
+        # fc1_weighted_output shape is (batch_size * sequence_len * topk, ffn_dim)
+        gate_up_output = gate_up_activation
+
+        down_output = group_gemm_same_nk(
+            a=gate_up_output,
+            b=down_proj,
+            cumsum_M=cumsum_t,
+            max_M=max_M,
+            transpose_a=False,
+            transpose_b=False,
+        )
+        down_output_bias = down_proj_bias.index_select(0, expert_index.flatten()).view(down_output.shape)
+        scattered_down_output_bias = torch.empty_like(down_output_bias)
+        scattered_down_output_bias[scatter_index.flatten()] = down_output_bias
+        down_output += scattered_down_output_bias
+
+        expert_output = moe_gather(down_output * scattered_gate_weight, scatter_index)  # 聚合 top-k 专家结果
+        output = expert_output.reshape(hidden_states.shape)
+    
+        # 保存反向传播所需张量
+        ctx.save_for_backward(
+            gate_weights, 
+            hidden_states, 
+            gate_up_proj,
+            gate_up_proj_bias,
+            down_proj,
+            down_proj_bias,
+            scatter_index,
+            scatter_output,
+            cumsum_t,
+            gate_up,
+            gate_up_output,
+            gate_up_activation,
+            scattered_gate_weight,
+            expert_index,
+            down_output,
+        )
+        ctx.alpha = alpha
+        ctx.limit = limit
+        ctx.num_experts = num_experts
+
+        return output
+
+    @staticmethod
+    def backward(ctx, grad_output):
+        # 恢复保存的张量和超参数
+        (
+            gate_weights, 
+            hidden_states, 
+            gate_up_proj,
+            gate_up_proj_bias,
+            down_proj,
+            down_proj_bias,
+            scatter_index,
+            scatter_output,
+            cumsum_t,
+            gate_up,
+            gate_up_output,
+            gate_up_activation,
+            scattered_gate_weight,
+            expert_index,
+            down_output,
+        ) = ctx.saved_tensors
+        alpha = ctx.alpha
+        limit = ctx.limit
+        num_experts = ctx.num_experts
+        hidden_dim = grad_output.shape[-1]
+        grad_output = grad_output.view(-1, hidden_dim)
+
+        # 反向传播核心逻辑（按前向步骤逆序）
+        grad_down_output = moe_scatter(grad_output, scatter_index)  # Step 7 逆过程
+
+        grad_scattered_gate_weight = torch.sum(grad_down_output * down_output, dim=-1)
+        grad_down_output *= scattered_gate_weight
+
+        # dgrad: 计算 gate_up_output 的梯度 (对应 down_proj 反向)
+        grad_gate_up_weight_output = group_gemm_same_nk(
+            a=grad_down_output,
+            b=down_proj,
+            cumsum_M=cumsum_t,
+            max_M=grad_output.shape[0],
+            transpose_b=True,
+        )
+
+        # wgrad: 计算 down_proj 的梯度 (对应 gate_up_output 反向)
+        grad_down_weight = None
+        if down_proj.requires_grad:
+            grad_down_weight = torch.empty_like(down_proj)
+            group_gemm_same_mn(
+                a=gate_up_output, # grad_down_output,
+                b=grad_down_output, # gate_up_output,
+                c=grad_down_weight,
+                cumsum_K=cumsum_t,
+                max_K=grad_output.shape[0],
+                transpose_a=True,
+                transpose_b=False,
+            )
+
+        # SwiGLU 激活反向传播（Step 5 逆过程）
+        # 2. 计算 gate 和 up 的梯度
+        gate = gate_up[..., ::2]
+        up = gate_up[..., 1::2]
+        up_clamp = up.clamp(min=-limit, max=limit) 
+        gate_clamped = gate.clamp(min=None, max=limit)
+        sigmoid_gate_alpha = torch.sigmoid(gate_clamped * alpha)
+
+        # 2.1 up 梯度 (含截断)
+        grad_up = grad_gate_up_weight_output * gate_clamped * sigmoid_gate_alpha
+        grad_up[up < -limit] = 0
+        grad_up[up > limit] = 0
+
+        # 2.2 gate 梯度 (含截断和 sigmoid 导数)
+        grad_gate_clamped = grad_gate_up_weight_output * (up_clamp + 1) * sigmoid_gate_alpha
+        grad_sigmoid = gate_clamped * grad_gate_up_weight_output * (up_clamp + 1)
+        grad_gate = grad_sigmoid * (sigmoid_gate_alpha * (1 - sigmoid_gate_alpha) * alpha) + grad_gate_clamped
+        grad_gate[gate > limit] = 0  # 截断区域梯度置零
+
+        # 3. 合并 gate/up 梯度并计算 gate_up_proj 梯度
+        grad_gate_up = torch.zeros_like(gate_up)
+        grad_gate_up[..., ::2] = grad_gate
+        grad_gate_up[..., 1::2] = grad_up
+
+        grad_scatter_output = group_gemm_same_nk(
+            a=grad_gate_up,
+            b=gate_up_proj,
+            cumsum_M=cumsum_t,
+            max_M=grad_output.shape[0],
+            transpose_b=True,
+        )
+
+        # 聚合梯度并恢复形状
+        grad_hidden_states = moe_gather(grad_scatter_output, scatter_index)
+        grad_hidden_states = grad_hidden_states.reshape(hidden_states.shape)
+
+        # 计算权重梯度（gate_up_proj 和 down_proj）
+        grad_gate_up_proj = None
+        if gate_up_proj.requires_grad:
+            grad_gate_up_proj = torch.empty_like(gate_up_proj)
+            group_gemm_same_mn(
+                a=scatter_output,
+                b=grad_gate_up,
+                c=grad_gate_up_proj,
+                cumsum_K=cumsum_t,
+                max_K=grad_output.shape[0],
+                transpose_a=True,
+                transpose_b=False,
+            )
+
+        # 偏置梯度（直接求和）
+        grad_gate_up_proj_bias = torch.zeros_like(gate_up_proj_bias)
+        grad_gate_up_proj_bias.index_add_(0, expert_index.flatten(), grad_gate_up[scatter_index.flatten()])
+
+        grad_down_proj_bias = torch.zeros_like(down_proj_bias)
+        grad_down_proj_bias.index_add_(0, expert_index.flatten(), grad_down_output[scatter_index.flatten()])
+
+        # 路由权重梯度
+        grad_gate_weights = grad_scattered_gate_weight[scatter_index.flatten()].reshape(gate_weights.shape)
+
+        return (
+            None,  # num_experts
+            grad_gate_weights,  # gate_weights
+            None,  # expert_index
+            grad_hidden_states,  # hidden_states
+            grad_gate_up_proj,  # gate_up_proj
+            grad_gate_up_proj_bias,  # gate_up_proj_bias
+            grad_down_weight,  # down_proj
+            grad_down_proj_bias,  # down_proj_bias
+            None,  # alpha（标量无梯度）
+            None,  # limit（标量无梯度）
+        )
+
+
+def fused_moe_forward(
+    module: torch.nn.Module,
+    num_experts: int,
+    routing_weights: torch.Tensor,
+    selected_experts: torch.Tensor,
+    hidden_states,
+    gate_up_proj,  # [num_experts, hidden, 2*ffn_dim]，交错布局
+    gate_up_proj_bias,
+    down_proj,  # [num_experts, ffn_dim, hidden]
+    down_proj_bias,
+    limit,
+    alpha,
+):
+    routing_weights = routing_weights.bfloat16()
+    hidden_states = hidden_states.bfloat16()
+    final_hidden_states = FusedMoeExpertFunction.apply(
+        num_experts,
+        routing_weights,
+        selected_experts,
+        hidden_states,
+        gate_up_proj,
+        gate_up_proj_bias,
+        down_proj,
+        down_proj_bias,
+        alpha,
+        limit,
+    )
+    return final_hidden_states