diff --git a/scripts/performance/configs/nemotronh/nemotron_3_nano_llm_pretrain.py b/scripts/performance/configs/nemotronh/nemotron_3_nano_llm_pretrain.py
index 4f23420451..3ecccd6f54 100644
--- a/scripts/performance/configs/nemotronh/nemotron_3_nano_llm_pretrain.py
+++ b/scripts/performance/configs/nemotronh/nemotron_3_nano_llm_pretrain.py
@@ -30,6 +30,8 @@ def set_nemotron_3_nano_common_configs(cfg: ConfigContainer) -> None:
     cfg.mixed_precision.grad_reduce_in_fp32 = False
     cfg.ddp.grad_reduce_in_fp32 = False
 
+    cfg.model.moe_router_force_load_balancing = True
+
 
 def nemotron_3_nano_pretrain_config_gb300(
     precision: str = "bf16", mock: bool = True, config_variant: str = "v1"
diff --git a/scripts/performance/configs/nemotronh/nemotron_3_nano_workload_base_configs.py b/scripts/performance/configs/nemotronh/nemotron_3_nano_workload_base_configs.py
index 7c85f88543..31310b3fea 100644
--- a/scripts/performance/configs/nemotronh/nemotron_3_nano_workload_base_configs.py
+++ b/scripts/performance/configs/nemotronh/nemotron_3_nano_workload_base_configs.py
@@ -41,6 +41,9 @@
 NEMOTRON_3_NANO_PRETRAIN_CONFIG_GB300_BF16_V1 = replace(
     BASE_NEMOTRON_3_NANO_CONFIG,
     tensor_model_parallel_size=1,
+    micro_batch_size=4,
+    cuda_graph_impl="transformer_engine",
+    cuda_graph_scope=["attn", "moe_router", "moe_preprocess", "mamba"],
 )
 NEMOTRON_3_NANO_PRETRAIN_CONFIG_GB300_FP8_MX_V1 = NEMOTRON_3_NANO_PRETRAIN_CONFIG_GB300_BF16_V1
 NEMOTRON_3_NANO_PRETRAIN_CONFIG_GB300_NVFP4_V1 = NEMOTRON_3_NANO_PRETRAIN_CONFIG_GB300_BF16_V1