diff --git a/.buildkite/cuda/test-nightly.yml b/.buildkite/cuda/test-nightly.yml
index ed112b84b..0c5d94d81 100644
--- a/.buildkite/cuda/test-nightly.yml
+++ b/.buildkite/cuda/test-nightly.yml
@@ -159,19 +159,7 @@ steps:
           - export VLLM_ALLOW_LONG_MAX_MODEL_LEN="1"
           - export VLLM_USE_DEEP_GEMM="0"
           - export VLLM_MOE_USE_DEEP_GEMM="0"
-          - pytest -s -v tests/e2e/ -m "full_model and L4 and tts" --run-level "full_model" --ignore=tests/e2e/accuracy --ignore=tests/e2e/online_serving/test_voxcpm2_tts_expansion.py
-        mirror_hardwares: l4_1
-
-      - label: ":full_moon: TTS · Accuracy Test · VoxCPM2"
-        timeout_in_minutes: 30
-        commands:
-          - |
-            timeout 30m bash -c "
-              export VLLM_LOGGING_LEVEL=DEBUG
-              export VLLM_WORKER_MULTIPROC_METHOD=spawn
-              export VLLM_ALLOW_LONG_MAX_MODEL_LEN=1
-              pytest -s -v tests/e2e/online_serving/test_voxcpm2_tts_expansion.py -m 'full_model and cuda and tts' --run-level 'full_model'
-            "
+          - pytest -s -v tests/e2e/ -m "full_model and L4 and tts" --run-level "full_model" --ignore=tests/e2e/accuracy
         mirror_hardwares: l4_1
 
       - label: ":full_moon: TTS · Perf Test"
@@ -188,35 +176,6 @@ steps:
             exit $$EXIT
         mirror_hardwares: h100_2
 
-      - label: ":full_moon: TTS · VoxCPM2 · Perf Test"
-        key: nightly-tts-performance-voxcpm2
-        timeout_in_minutes: 180
-        commands:
-          - export BENCHMARK_DIR=tests/dfx/perf/results
-          - export VLLM_ALLOW_LONG_MAX_MODEL_LEN="1"
-          - |
-            set +e
-            pytest -s -v tests/dfx/perf/scripts/run_benchmark.py --test-config-file tests/dfx/perf/tests/test_voxcpm2.json
-            EXIT=$$?
-            buildkite-agent artifact upload "tests/dfx/perf/results/*.json"
-            exit $$EXIT
-        mirror_hardwares: h100_1
-
-      - label: ":full_moon: TTS · Higgs-Audio-V3 · Perf Test"
-        key: nightly-tts-performance-higgs-audio-v3
-        timeout_in_minutes: 180
-        commands:
-          - export BENCHMARK_DIR=tests/dfx/perf/results
-          - export VLLM_USE_DEEP_GEMM="0"
-          - export VLLM_MOE_USE_DEEP_GEMM="0"
-          - |
-            set +e
-            pytest -s -v tests/dfx/perf/scripts/run_benchmark.py --test-config-file tests/dfx/perf/tests/test_higgs_audio_v3.json
-            EXIT=$$?
-            buildkite-agent artifact upload "tests/dfx/perf/results/*.json"
-            exit $$EXIT
-        mirror_hardwares: h100_1
-
   - group: ":card_index_dividers: Tiny Model Tests [Multi-GPU]"
     key: nightly-tiny-model-test-group
     depends_on: upload-nightly-pipeline
@@ -246,8 +205,7 @@ steps:
       build.pull_request.labels includes "nightly-test" ||
       build.pull_request.labels includes "diffusion-x2iat-test"
     steps:
-      # Use explicit file shards so single-card cases do not reserve 2+ GPUs,
-      # and Bagel feature coverage is not duplicated via a broad tests/e2e sweep.
+      # Use explicit file shards so single-card cases do not reserve 2+ GPUs.
 
       - label: ":full_moon: Diffusion X2I(&A&T) · Function Test with H100 · Single-GPU"
         timeout_in_minutes: 120
@@ -255,15 +213,6 @@ steps:
           - >-
             pytest -sv
             tests/e2e/online_serving/test_qwen_image_expansion.py
-            tests/e2e/online_serving/test_qwen_image_edit_expansion.py
-            tests/e2e/online_serving/test_qwen_image_layered_expansion.py
-            tests/e2e/online_serving/test_boogu_image.py
-            tests/e2e/online_serving/test_boogu_image_edit.py
-            tests/e2e/online_serving/test_boogu_image_expansion.py
-            tests/e2e/online_serving/test_boogu_image_edit_expansion.py
-            tests/e2e/online_serving/test_bagel_expansion.py
-            tests/e2e/online_serving/test_krea2_expansion.py
-            tests/e2e/offline_inference/test_krea2_expansion.py
             -m "full_model and diffusion and H100 and not distributed_cuda"
             --run-level "full_model"
         mirror_hardwares: h100_1
@@ -274,47 +223,14 @@ steps:
           - >-
             pytest -sv
             tests/e2e/online_serving/test_qwen_image_expansion.py
-            tests/e2e/online_serving/test_qwen_image_edit_expansion.py
-            tests/e2e/offline_inference/test_krea2_expansion.py
             -m "full_model and diffusion and H100 and distributed_cuda"
             --run-level "full_model"
         mirror_hardwares: h100_2
 
-      - label: ":full_moon: Diffusion X2I(&A&T) · Function Test with H100 · Multi-GPU Layered/LongCat/FLUX"
-        timeout_in_minutes: 90
-        commands:
-          - >-
-            pytest -sv
-            tests/e2e/online_serving/test_qwen_image_layered_expansion.py
-            -m "full_model and diffusion and H100 and distributed_cuda"
-            --run-level "full_model"
-        mirror_hardwares: h100_2
-
-      - label: ":full_moon: Diffusion X2I(&A&T) · Function Test with H100 · Multi-GPU Bagel"
-        timeout_in_minutes: 120
-        commands:
-          - >-
-            pytest -sv
-            tests/e2e/online_serving/test_bagel_expansion.py
-            -m "full_model and diffusion and H100 and distributed_cuda"
-            --run-level "full_model"
-        mirror_hardwares: h100_2
-
-      - label: ":full_moon: Diffusion X2I(&A&T) · GLM-Image Function Test with H100"
-        timeout_in_minutes: 120
-        commands:
-          - >-
-            pytest -sv
-            tests/e2e/online_serving/test_glm_image_expansion.py
-            tests/e2e/offline_inference/test_glm_image_autoround_w4a16_expansion.py
-            -m "full_model and diffusion and H100"
-            --run-level "full_model"
-        mirror_hardwares: h100_2
-
       - label: ":full_moon: Diffusion X2I(&A&T) · Function Test with L4"
         timeout_in_minutes: 120
         commands:
-          - pytest -sv tests/e2e/ -k "not test_wan and not test_bagel_expansion and not hunyuan" -m "full_model and diffusion and L4" --run-level "full_model" --ignore=tests/e2e/accuracy
+          - pytest -sv tests/e2e/ -k "not test_wan and not hunyuan" -m "full_model and diffusion and L4" --run-level "full_model" --ignore=tests/e2e/accuracy
         mirror_hardwares: l4_4
 
       - label: ":full_moon: Diffusion X2I(&A&T) · Doc Test"
@@ -327,7 +243,7 @@ steps:
         timeout_in_minutes: 180
         commands:
           - export VLLM_HTTP_TIMEOUT_KEEP_ALIVE=120
-          - pytest -s -v tests/e2e/accuracy/test_qwen_image*.py --run-level full_model
+          - pytest -s -v tests/e2e/accuracy/test_qwen_image.py --run-level full_model
           - pytest -s -v tests/e2e/accuracy/test_diffusers_backend_similarity.py -k '2i_matches_diffusers' -m full_model --run-level full_model
         mirror_hardwares: h100_1
 
@@ -355,70 +271,6 @@ steps:
             exit $$EXIT
         mirror_hardwares: h100_4
 
-      - label: ":full_moon: Diffusion X2I(&A&T) · Perf Test · Qwen-Image-Edit"
-        key: nightly-diffusion-x2iat-performance-qwen-image-edit
-        timeout_in_minutes: 180
-        commands:
-          - export DIFFUSION_BENCHMARK_DIR=tests/dfx/perf/results
-          - export DIFFUSION_ATTENTION_BACKEND=FLASH_ATTN
-          - export CACHE_DIT_VERSION=1.5.0
-          - |
-            set +e
-            pytest -s -v tests/dfx/perf/scripts/run_diffusion_benchmark.py --test-config-file tests/dfx/perf/tests/test_qwen_image_edit_vllm_omni.json
-            EXIT=$$?
-            buildkite-agent artifact upload "tests/dfx/perf/results/diffusion_result_*.json"
-            buildkite-agent artifact upload "tests/dfx/perf/results/logs/*.log"
-            exit $$EXIT
-        mirror_hardwares: h100_4
-
-      - label: ":full_moon: Diffusion X2I(&A&T) · Perf Test · Qwen-Image-Edit-2511"
-        key: nightly-diffusion-x2iat-performance-qwen-image-edit-2511
-        timeout_in_minutes: 180
-        commands:
-          - export DIFFUSION_BENCHMARK_DIR=tests/dfx/perf/results
-          - export DIFFUSION_ATTENTION_BACKEND=FLASH_ATTN
-          - export CACHE_DIT_VERSION=1.5.0
-          - |
-            set +e
-            pytest -s -v tests/dfx/perf/scripts/run_diffusion_benchmark.py --test-config-file tests/dfx/perf/tests/test_qwen_image_edit_2511_vllm_omni.json
-            EXIT=$$?
-            buildkite-agent artifact upload "tests/dfx/perf/results/diffusion_result_*.json"
-            buildkite-agent artifact upload "tests/dfx/perf/results/logs/*.log"
-            exit $$EXIT
-        mirror_hardwares: h100_4
-
-      - label: ":full_moon: Diffusion X2I(&A&T) · Perf Test · Qwen-Image-Layered"
-        key: nightly-diffusion-x2iat-performance-qwen-image-layered
-        timeout_in_minutes: 120
-        commands:
-          - export DIFFUSION_BENCHMARK_DIR=tests/dfx/perf/results
-          - export DIFFUSION_ATTENTION_BACKEND=FLASH_ATTN
-          - export CACHE_DIT_VERSION=1.5.0
-          - |
-            set +e
-            pytest -s -v tests/dfx/perf/scripts/run_diffusion_benchmark.py --test-config-file tests/dfx/perf/tests/test_qwen_image_layered_vllm_omni.json
-            EXIT=$$?
-            buildkite-agent artifact upload "tests/dfx/perf/results/diffusion_result_*.json"
-            buildkite-agent artifact upload "tests/dfx/perf/results/logs/*.log"
-            exit $$EXIT
-        mirror_hardwares: h100_1
-
-      - label: ":full_moon: Diffusion X2I(&A&T) · Perf Test · BAGEL"
-        key: nightly-diffusion-x2iat-performance-bagel
-        timeout_in_minutes: 180
-        commands:
-          - export DIFFUSION_BENCHMARK_DIR=tests/dfx/perf/results
-          - export DIFFUSION_ATTENTION_BACKEND=FLASH_ATTN
-          - export CACHE_DIT_VERSION=1.5.0
-          - |
-            set +e
-            pytest -s -v tests/dfx/perf/scripts/run_diffusion_benchmark.py --test-config-file tests/dfx/perf/tests/test_bagel_vllm_omni.json
-            EXIT=$$?
-            buildkite-agent artifact upload "tests/dfx/perf/results/diffusion_result_*.json"
-            buildkite-agent artifact upload "tests/dfx/perf/results/logs/*.log"
-            exit $$EXIT
-        mirror_hardwares: h100_1
-
   - group: ":card_index_dividers: Diffusion X2V Model Test"
     key: nightly-diffusion-x2v-group
     depends_on: upload-nightly-pipeline
@@ -440,12 +292,6 @@ steps:
           - pytest -s -v tests/e2e/online_serving/test_wan22_expansion.py -k "i2v" -m "full_model and cuda" --run-level "full_model"
         mirror_hardwares: h100_2
 
-      - label: ":full_moon: Diffusion X2V · LingBot Video MoE Function Test"
-        timeout_in_minutes: 90
-        commands:
-          - pytest -s -v tests/e2e/online_serving/test_lingbot_video_moe.py -k "video_generation" -m "full_model and cuda" --run-level "full_model"
-        mirror_hardwares: h100_1
-
       - label: ":full_moon: Diffusion X2V · Doc Test"
         timeout_in_minutes: 60
         commands:
@@ -456,15 +302,8 @@ steps:
         timeout_in_minutes: 90
         commands:
           - pytest -s -v tests/e2e/offline_inference/test_wan22_autoround_w4a16_expansion.py -m "full_model and cuda" --run-level "full_model"
-          - pytest -s -v tests/e2e/online_serving/test_wan_2_1_vace_expansion.py tests/e2e/online_serving/test_ltx.py -m "full_model and cuda" --run-level "full_model"
         mirror_hardwares: h100_2
 
-      - label: ":full_moon: Diffusion X2V · LTX-2.5 Online Test"
-        timeout_in_minutes: 60
-        commands:
-          - pytest -s -v tests/e2e/online_serving/test_ltx25.py -m "full_model and cuda" --run-level "full_model"
-        mirror_hardwares: h100_1
-
       - label: ":full_moon: Diffusion X2V · Wan2.2 Accuracy Test"
         timeout_in_minutes: 180
         commands:
@@ -483,23 +322,8 @@ steps:
         commands:
           - pytest -s -v tests/e2e/accuracy/hunyuanvideo15_t2v/test_hunyuanvideo15_t2v_video_similarity.py -m full_model --run-level full_model
           - pytest -s -v tests/e2e/accuracy/hunyuanvideo15_i2v/test_hunyuanvideo15_i2v_video_similarity.py -m full_model --run-level full_model
-          - pytest -s -v tests/e2e/accuracy/test_video_streaming_output_similarity.py -m full_model --run-level full_model
         mirror_hardwares: h100_2
 
-      - label: ":full_moon: Diffusion X2V · LTX Accuracy Test"
-        timeout_in_minutes: 180
-        commands:
-          - pytest -s -v 'tests/e2e/accuracy/ltx/test_ltx_official_similarity.py::test_ltx_matches_official[ltx2_3_i2v]' -m full_model --run-level full_model
-          - pytest -s -v 'tests/e2e/accuracy/ltx/test_ltx_official_similarity.py::test_ltx_matches_official[ltx23_two_stage_layer_fused_i2v]' -m full_model --run-level full_model
-        mirror_hardwares: h100_1
-
-      - label: ":full_moon: Diffusion X2V · LTX-2.5 Official Parity Test"
-        timeout_in_minutes: 180
-        commands:
-          - pytest -s -v tests/e2e/accuracy/ltx/test_ltx25_official_similarity.py::test_ltx25_distilled_two_stage_matches_official -m full_model --run-level full_model
-          - pytest -s -v tests/e2e/accuracy/ltx/test_ltx25_official_similarity.py::test_ltx25_full_matches_official -m full_model --run-level full_model
-        mirror_hardwares: h100_1
-
       - label: ":full_moon: Diffusion X2V · Perf Test"
         key: nightly-diffusion-x2v-performance
         timeout_in_minutes: 180
@@ -575,16 +399,6 @@ steps:
           - pytest -sv tests/diffusion/quantization -m 'full_model and cuda and L4' --run-level "full_model"
         mirror_hardwares: l4_1
 
-  - label: ":bar_chart: Testcase Statistics"
-    key: nightly-testcase-statistics
-    timeout_in_minutes: 120
-    depends_on: upload-nightly-pipeline
-    if: build.env("NIGHTLY") == "1"
-    commands:
-      - python tools/nightly/buildkite_testcase_statistics.py -o tests/dfx/perf/results/buildkite_testcase_statistics.html
-      - buildkite-agent artifact upload "tests/dfx/perf/results/*.html"
-    mirror_hardwares: h100_1
-
   # NOTE: ":email: Nightly Collection & Email" is deprecated and will be removed soon.
   # Do not add new depends_on entries or extend this step.
   - label: ":email: Nightly Collection & Email"
@@ -595,32 +409,20 @@ steps:
       - nightly-omni-performance-vllm-text
       - nightly-omni-performance-multi-replicas
       - nightly-tts-performance
-      - nightly-tts-performance-higgs-audio-v3
       - nightly-diffusion-x2iat-performance-qwen-image
-      - nightly-diffusion-x2iat-performance-qwen-image-edit
-      - nightly-diffusion-x2iat-performance-qwen-image-edit-2511
-      - nightly-diffusion-x2iat-performance-qwen-image-layered
-      - nightly-diffusion-x2iat-performance-bagel
       - nightly-diffusion-x2v-performance
-      - nightly-testcase-statistics
     if: build.env("NIGHTLY") == "1" && build.env("EMAIL_DISTRIBUTION") == "1"
     commands:
       - pip install openpyxl
       - export DEFAULT_INPUT_DIR=tests/dfx/perf/results
       - export DEFAULT_OUTPUT_DIR=tests/dfx/perf/results
       - buildkite-agent artifact download "tests/dfx/perf/results/*.json" . --step nightly-tts-performance
-      - buildkite-agent artifact download "tests/dfx/perf/results/*.json" . --step nightly-tts-performance-higgs-audio-v3
       - buildkite-agent artifact download "tests/dfx/perf/results/*.json" . --step nightly-omni-performance-no-async-chunk
       - buildkite-agent artifact download "tests/dfx/perf/results/*.json" . --step nightly-omni-performance-async-chunk
       - buildkite-agent artifact download "tests/dfx/perf/results/*.json" . --step nightly-omni-performance-vllm-text
       - buildkite-agent artifact download "tests/dfx/perf/results/*.json" . --step nightly-omni-performance-multi-replicas
       - buildkite-agent artifact download "tests/dfx/perf/results/*.json" . --step nightly-diffusion-x2iat-performance-qwen-image
-      - buildkite-agent artifact download "tests/dfx/perf/results/*.json" . --step nightly-diffusion-x2iat-performance-qwen-image-edit
-      - buildkite-agent artifact download "tests/dfx/perf/results/*.json" . --step nightly-diffusion-x2iat-performance-qwen-image-edit-2511
-      - buildkite-agent artifact download "tests/dfx/perf/results/*.json" . --step nightly-diffusion-x2iat-performance-qwen-image-layered
-      - buildkite-agent artifact download "tests/dfx/perf/results/*.json" . --step nightly-diffusion-x2iat-performance-bagel
       - buildkite-agent artifact download "tests/dfx/perf/results/*.json" . --step nightly-diffusion-x2v-performance
-      - buildkite-agent artifact download "tests/dfx/perf/results/*.html" . --step nightly-testcase-statistics
       - python tools/nightly/generate_nightly_perf_excel.py
       - python tools/nightly/generate_nightly_perf_html.py
       - python tools/nightly/send_nightly_email.py --report-file "tests/dfx/perf/results/*.xlsx, tests/dfx/perf/results/*.html"
diff --git a/.buildkite/cuda/test-weekly.yml b/.buildkite/cuda/test-weekly.yml
index 0fd3a06ce..5ddbf22ce 100644
--- a/.buildkite/cuda/test-weekly.yml
+++ b/.buildkite/cuda/test-weekly.yml
@@ -149,4 +149,4 @@ steps:
             pytest -sv
             tests/e2e/ -m "slow and diffusion and H100 and distributed_cuda"
             --run-level "full_model"
-        mirror_hardwares: h100_2
+        mirror_hardwares: h100_4
diff --git a/.buildkite/npu/test-npu-nightly.yml b/.buildkite/npu/test-npu-nightly.yml
index 050cef4f7..2cdf0cd54 100644
--- a/.buildkite/npu/test-npu-nightly.yml
+++ b/.buildkite/npu/test-npu-nightly.yml
@@ -96,33 +96,6 @@ steps:
             EXIT=$$?
             buildkite-agent artifact upload "tests/dfx/perf/results/*.json"
             exit $$EXIT
-  - group: ":card_index_dividers: Diffusion X2I(&A&T) Model Test"
-    key: nightly-diffusion-x2iat-group
-    depends_on: upload-nightly-pipeline
-    if: >-
-      build.env("NIGHTLY") == "1" ||
-      build.pull_request.labels includes "nightly-test" ||
-      build.pull_request.labels includes "diffusion-x2iat-test"
-    steps:
-      - label: ":full_moon: Diffusion X2I(&A&T) · Perf Test · Qwen-Image-Edit-2511"
-        key: nightly-diffusion-x2iat-performance-qwen-image-edit-2511
-        timeout_in_minutes: 180
-        mirror_hardwares: a3_npu_8
-        env:
-          VLLM_WORKER_MULTIPROC_METHOD: spawn
-          VLLM_ALLOW_LONG_MAX_MODEL_LEN: "1"
-          HF_TOKEN: "${HF_TOKEN}"
-          DIFFUSION_ATTENTION_BACKEND: TORCH_SDPA
-        commands:
-          - export DIFFUSION_BENCHMARK_DIR=tests/dfx/perf/results
-          - export CACHE_DIT_VERSION=1.5.0
-          - |
-            set +e
-            pytest -s -v tests/dfx/perf/scripts/run_diffusion_benchmark.py -m npu --test-config-file tests/dfx/perf/tests/test_qwen_image_edit_2511_vllm_omni.json
-            EXIT=$$?
-            buildkite-agent artifact upload "tests/dfx/perf/results/diffusion_result_*.json"
-            buildkite-agent artifact upload "tests/dfx/perf/results/logs/*.log"
-            exit $$EXIT
   - group: ":card_index_dividers: Diffusion I2I Model Test"
     key: nightly-HunyuanImage3-x2i-test-group
     depends_on: upload-nightly-pipeline
diff --git a/tests/dfx/perf/tests/test_bagel_vllm_omni.json b/tests/dfx/perf/tests/test_bagel_vllm_omni.json
index 9f6ff0241..9dda3148d 100644
--- a/tests/dfx/perf/tests/test_bagel_vllm_omni.json
+++ b/tests/dfx/perf/tests/test_bagel_vllm_omni.json
@@ -10,7 +10,7 @@
           "num_cards": 1
         }
       },
-      "full_model",
+      "local_model",
       "diffusion"
     ],
     "description": "Single-stage BAGEL (TP=1, CFG=1), t2i Task",
@@ -61,7 +61,7 @@
           "num_cards": 1
         }
       },
-      "full_model",
+      "local_model",
       "diffusion"
     ],
     "description": "Single-stage BAGEL (TP=1, CFG=1), i2i Task",
@@ -113,7 +113,7 @@
           "num_cards": 1
         }
       },
-      "full_model",
+      "local_model",
       "diffusion"
     ],
     "description": "Two-stage BAGEL (bagel.yaml), TP=1, CFG=1, t2i Task",
@@ -164,7 +164,7 @@
           "num_cards": 1
         }
       },
-      "full_model",
+      "local_model",
       "diffusion"
     ],
     "description": "Two-stage BAGEL (bagel.yaml), TP=1, CFG=1, i2i Task",
diff --git a/tests/dfx/perf/tests/test_higgs_audio_v3.json b/tests/dfx/perf/tests/test_higgs_audio_v3.json
index 75e5983b0..1197c88f1 100644
--- a/tests/dfx/perf/tests/test_higgs_audio_v3.json
+++ b/tests/dfx/perf/tests/test_higgs_audio_v3.json
@@ -10,7 +10,7 @@
           "num_cards": 1
         }
       },
-      "full_model",
+      "local_model",
       "tts"
     ],
     "server_params": {
diff --git a/tests/dfx/perf/tests/test_ltx2_vllm_omni.json b/tests/dfx/perf/tests/test_ltx2_vllm_omni.json
index 1a0f31e98..9b0634a20 100644
--- a/tests/dfx/perf/tests/test_ltx2_vllm_omni.json
+++ b/tests/dfx/perf/tests/test_ltx2_vllm_omni.json
@@ -10,7 +10,7 @@
           "num_cards": 1
         }
       },
-      "full_model",
+      "local_model",
       "diffusion"
     ],
     "description": "Single-device baseline with enforce-eager (no torch.compile)",
@@ -64,7 +64,7 @@
           "num_cards": 1
         }
       },
-      "full_model",
+      "local_model",
       "diffusion"
     ],
     "description": "Single-device with torch.compile (default, no enforce-eager)",
@@ -117,7 +117,7 @@
           "num_cards": 2
         }
       },
-      "full_model",
+      "local_model",
       "diffusion"
     ],
     "description": "CFG-parallel=2 with enforce-eager",
@@ -172,7 +172,7 @@
           "num_cards": 2
         }
       },
-      "full_model",
+      "local_model",
       "diffusion"
     ],
     "description": "CFG-parallel=2 with torch.compile",
@@ -226,7 +226,7 @@
           "num_cards": 1
         }
       },
-      "full_model",
+      "local_model",
       "diffusion"
     ],
     "description": "CacheDiT with enforce-eager",
diff --git a/tests/dfx/perf/tests/test_qwen_image_edit_2511_vllm_omni.json b/tests/dfx/perf/tests/test_qwen_image_edit_2511_vllm_omni.json
index dcd2188c5..fd8e437fd 100644
--- a/tests/dfx/perf/tests/test_qwen_image_edit_2511_vllm_omni.json
+++ b/tests/dfx/perf/tests/test_qwen_image_edit_2511_vllm_omni.json
@@ -10,7 +10,7 @@
           "num_cards": 1
         }
       },
-      "full_model",
+      "local_model",
       "diffusion"
     ],
     "description": "Single-device baseline (two input images)",
@@ -76,7 +76,7 @@
           "num_cards": 4
         }
       },
-      "full_model",
+      "local_model",
       "diffusion"
     ],
     "description": "Ulysses SP=2 + CFG=2 + VAE patch parallel=4",
@@ -125,7 +125,7 @@
           "num_cards": 4
         }
       },
-      "full_model",
+      "local_model",
       "diffusion"
     ],
     "description": "Ulysses SP=2 + CFG=2 + CacheDiT",
diff --git a/tests/dfx/perf/tests/test_qwen_image_edit_vllm_omni.json b/tests/dfx/perf/tests/test_qwen_image_edit_vllm_omni.json
index 686bd6160..62ee70aa2 100644
--- a/tests/dfx/perf/tests/test_qwen_image_edit_vllm_omni.json
+++ b/tests/dfx/perf/tests/test_qwen_image_edit_vllm_omni.json
@@ -10,7 +10,7 @@
           "num_cards": 1
         }
       },
-      "full_model",
+      "local_model",
       "diffusion"
     ],
     "description": "Single-device baseline",
@@ -73,7 +73,7 @@
           "num_cards": 4
         }
       },
-      "full_model",
+      "local_model",
       "diffusion"
     ],
     "description": "Ulysses SP=2 + CFG=2 + VAE patch parallel=4",
@@ -121,7 +121,7 @@
           "num_cards": 4
         }
       },
-      "full_model",
+      "local_model",
       "diffusion"
     ],
     "description": "Ulysses SP=2 + CFG=2 + CacheDiT",
diff --git a/tests/dfx/perf/tests/test_qwen_image_layered_vllm_omni.json b/tests/dfx/perf/tests/test_qwen_image_layered_vllm_omni.json
index bc4d93a24..b5fa2815b 100644
--- a/tests/dfx/perf/tests/test_qwen_image_layered_vllm_omni.json
+++ b/tests/dfx/perf/tests/test_qwen_image_layered_vllm_omni.json
@@ -10,7 +10,7 @@
           "num_cards": 1
         }
       },
-      "full_model",
+      "local_model",
       "diffusion"
     ],
     "description": "Single-device baseline",
diff --git a/tests/dfx/perf/tests/test_voxcpm2.json b/tests/dfx/perf/tests/test_voxcpm2.json
index e5163e5bb..1756bc206 100644
--- a/tests/dfx/perf/tests/test_voxcpm2.json
+++ b/tests/dfx/perf/tests/test_voxcpm2.json
@@ -10,7 +10,7 @@
           "num_cards": 1
         }
       },
-      "full_model",
+      "local_model",
       "tts"
     ],
     "server_params": {
diff --git a/tests/e2e/accuracy/ltx/test_ltx25_official_similarity.py b/tests/e2e/accuracy/ltx/test_ltx25_official_similarity.py
index 9f1623c8a..d72d14022 100644
--- a/tests/e2e/accuracy/ltx/test_ltx25_official_similarity.py
+++ b/tests/e2e/accuracy/ltx/test_ltx25_official_similarity.py
@@ -463,7 +463,7 @@ def _audio_metrics(reference: np.ndarray, prediction: np.ndarray) -> dict[str, f
     }
 
 
-@pytest.mark.full_model
+@pytest.mark.slow
 @pytest.mark.benchmark
 @pytest.mark.diffusion
 @hardware_test(res={"cuda": "H100"}, num_cards=1)
@@ -591,7 +591,7 @@ def test_ltx25_distilled_two_stage_matches_official(accuracy_artifact_root: Path
     assert audio_metrics["relative_l2"] <= LTX25_AUDIO_RELATIVE_L2_THRESHOLD
 
 
-@pytest.mark.full_model
+@pytest.mark.slow
 @pytest.mark.benchmark
 @pytest.mark.diffusion
 @hardware_test(res={"cuda": "H100"}, num_cards=1)
diff --git a/tests/e2e/accuracy/ltx/test_ltx_official_similarity.py b/tests/e2e/accuracy/ltx/test_ltx_official_similarity.py
index 8a3be4a99..59a8540aa 100644
--- a/tests/e2e/accuracy/ltx/test_ltx_official_similarity.py
+++ b/tests/e2e/accuracy/ltx/test_ltx_official_similarity.py
@@ -544,7 +544,7 @@ def _audio_metrics(reference: np.ndarray, prediction: np.ndarray) -> dict[str, f
 
 
 @pytest.mark.parametrize("case", CASES, ids=lambda case: case.name)
-@pytest.mark.full_model
+@pytest.mark.slow
 @pytest.mark.benchmark
 @pytest.mark.diffusion
 @hardware_test(res={"cuda": "H100"}, num_cards=1)
diff --git a/tests/e2e/accuracy/test_qwen_image_edit.py b/tests/e2e/accuracy/test_qwen_image_edit.py
index 9ee9c2c87..935a032d8 100644
--- a/tests/e2e/accuracy/test_qwen_image_edit.py
+++ b/tests/e2e/accuracy/test_qwen_image_edit.py
@@ -17,7 +17,7 @@ from tests.helpers.mark import hardware_test
 from tests.helpers.media import get_asset_path
 from tests.helpers.runtime import OmniServer
 
-pytestmark = [pytest.mark.full_model, pytest.mark.diffusion]
+pytestmark = [pytest.mark.diffusion]
 
 
 SINGLE_MODEL = "Qwen/Qwen-Image-Edit"
@@ -253,6 +253,7 @@ def _diffusers_output_single_image(accuracy_artifact_root: Path, qwen_bear_image
     )
 
 
+@pytest.mark.slow
 @pytest.mark.benchmark
 @hardware_test(res={"cuda": "H100"}, num_cards=1)
 def test_qwen_image_edit_single_matches_diffusers(
@@ -278,6 +279,7 @@ def test_qwen_image_edit_single_matches_diffusers(
     )
 
 
+@pytest.mark.slow
 @pytest.mark.benchmark
 @hardware_test(res={"cuda": "H100"}, num_cards=1)
 def test_qwen_image_edit_2511_matches_diffusers_pixelwise(accuracy_artifact_root: Path) -> None:
diff --git a/tests/e2e/accuracy/test_qwen_image_layered.py b/tests/e2e/accuracy/test_qwen_image_layered.py
index f2bc1b6aa..3ad8d3328 100644
--- a/tests/e2e/accuracy/test_qwen_image_layered.py
+++ b/tests/e2e/accuracy/test_qwen_image_layered.py
@@ -17,7 +17,7 @@ from tests.helpers.env import run_post_test_cleanup, run_pre_test_cleanup
 from tests.helpers.mark import hardware_test
 from tests.helpers.runtime import OmniServer
 
-pytestmark = [pytest.mark.full_model, pytest.mark.diffusion]
+pytestmark = [pytest.mark.slow, pytest.mark.diffusion]
 
 
 MODEL_ID = "Qwen/Qwen-Image-Layered"
diff --git a/tests/e2e/accuracy/test_video_streaming_output_similarity.py b/tests/e2e/accuracy/test_video_streaming_output_similarity.py
index 799e6e125..1aa7cdfc2 100644
--- a/tests/e2e/accuracy/test_video_streaming_output_similarity.py
+++ b/tests/e2e/accuracy/test_video_streaming_output_similarity.py
@@ -11,7 +11,7 @@ from tests.e2e.accuracy.helpers import assert_video_similarity_metrics, probe_vi
 from tests.helpers.mark import hardware_marks
 from tests.helpers.runtime import DiffusionResponse, OmniServer, OpenAIClientHandler
 
-pytestmark = [pytest.mark.diffusion, pytest.mark.full_model]
+pytestmark = [pytest.mark.diffusion, pytest.mark.slow]
 
 MODEL = "BestWishYsh/Helios-Distilled"
 PROMPT = "A serene lakeside sunrise with mist over the water."
diff --git a/tests/e2e/offline_inference/test_bagel_expansion.py b/tests/e2e/offline_inference/test_bagel_expansion.py
index e8ac6c19f..4c31be722 100644
--- a/tests/e2e/offline_inference/test_bagel_expansion.py
+++ b/tests/e2e/offline_inference/test_bagel_expansion.py
@@ -138,7 +138,7 @@ def _make_file_lora_request(adapter_dir: Path) -> LoRARequest:
 
 
 pytestmark = [
-    pytest.mark.full_model,
+    pytest.mark.slow,
     pytest.mark.diffusion,
     pytest.mark.parametrize("omni_runner", [_OMNI_RUNNER_PARAM], indirect=True),
 ]
diff --git a/tests/e2e/offline_inference/test_glm_image_autoround_w4a16_expansion.py b/tests/e2e/offline_inference/test_glm_image_autoround_w4a16_expansion.py
index 28eaab9b3..86b8cee8a 100644
--- a/tests/e2e/offline_inference/test_glm_image_autoround_w4a16_expansion.py
+++ b/tests/e2e/offline_inference/test_glm_image_autoround_w4a16_expansion.py
@@ -109,7 +109,7 @@ def _diffusion_sampling_params(
 
 
 pytestmark = [
-    pytest.mark.full_model,
+    pytest.mark.slow,
     pytest.mark.diffusion,
 ]
 
diff --git a/tests/e2e/offline_inference/test_moss_tts_realtime.py b/tests/e2e/offline_inference/test_moss_tts_realtime.py
index 892a395ba..34d36118d 100644
--- a/tests/e2e/offline_inference/test_moss_tts_realtime.py
+++ b/tests/e2e/offline_inference/test_moss_tts_realtime.py
@@ -96,7 +96,7 @@ def _get_test_config() -> str:
 
 pytestmark = [
     pytest.mark.skip(reason="https://github.com/vllm-project/vllm-omni/issues/4700"),
-    pytest.mark.full_model,
+    pytest.mark.slow,
     pytest.mark.tts,
     pytest.mark.parametrize(
         "omni_runner",
@@ -271,7 +271,6 @@ def _collect_audio(omni_runner: OmniRunner, request: dict) -> tuple[torch.Tensor
 # ---------------------------------------------------------------------------
 
 
-@pytest.mark.advanced_model
 @hardware_test(res={"cuda": "L4"}, num_cards=1)
 def test_moss_tts_realtime_english(omni_runner: OmniRunner, ref_audio_path: str) -> None:
     """MossTTSRealtime: English voice_clone produces non-empty 24 kHz audio."""
@@ -283,7 +282,6 @@ def test_moss_tts_realtime_english(omni_runner: OmniRunner, ref_audio_path: str)
     assert not torch.all(audio == 0), "Audio is silence"
 
 
-@pytest.mark.advanced_model
 @hardware_test(res={"cuda": "L4"}, num_cards=1)
 def test_moss_tts_realtime_chinese(omni_runner: OmniRunner, ref_audio_path: str) -> None:
     """MossTTSRealtime: Chinese input produces non-empty audio."""
diff --git a/tests/e2e/online_serving/test_bagel_expansion.py b/tests/e2e/online_serving/test_bagel_expansion.py
index fa17791e5..3308efdcb 100644
--- a/tests/e2e/online_serving/test_bagel_expansion.py
+++ b/tests/e2e/online_serving/test_bagel_expansion.py
@@ -22,7 +22,7 @@ from tests.helpers.mark import hardware_marks
 from tests.helpers.runtime import OmniServer, OmniServerParams, OpenAIClientHandler, dummy_messages_from_mix_data
 from tests.helpers.stage_config import get_deploy_config_path, modify_stage_config
 
-pytestmark = [pytest.mark.diffusion, pytest.mark.full_model]
+pytestmark = [pytest.mark.diffusion, pytest.mark.slow]
 
 PROMPT = "A futuristic city skyline at twilight, cyberpunk style, ultra-detailed, high resolution."
 NEGATIVE_PROMPT = "low quality, blurry, distorted, deformed, watermark"
diff --git a/tests/e2e/online_serving/test_bagel_multi_replicas.py b/tests/e2e/online_serving/test_bagel_multi_replicas.py
index cb7aeeb94..6409ddd58 100644
--- a/tests/e2e/online_serving/test_bagel_multi_replicas.py
+++ b/tests/e2e/online_serving/test_bagel_multi_replicas.py
@@ -1,6 +1,8 @@
-"""Nightly-only e2e coverage for BAGEL diffusion multi-replica serving.
+"""Weekly e2e coverage for BAGEL diffusion multi-replica serving.
 
 This test needs 4 H100 GPUs; keep it out of generic test_bagel_* jobs.
+It is covered by the weekly Diffusion · H100 · Multi-GPU lane
+(``.buildkite/cuda/test-weekly.yml``, ``mirror_hardwares: h100_4``).
 """
 
 import os
@@ -17,7 +19,7 @@ MODEL = "ByteDance-Seed/BAGEL-7B-MoT"
 MULTI_REPLICA_DEPLOY = get_deploy_config_path("ci/bagel_multi_replicas_4gpu.yaml")
 ROUTE_STRESS_REQUESTS = 6
 
-pytestmark = [pytest.mark.full_model, pytest.mark.diffusion]
+pytestmark = [pytest.mark.slow, pytest.mark.diffusion]
 
 test_params = [
     OmniServerParams(
diff --git a/tests/e2e/online_serving/test_boogu_image.py b/tests/e2e/online_serving/test_boogu_image.py
index 5a5494d7d..440e031f1 100644
--- a/tests/e2e/online_serving/test_boogu_image.py
+++ b/tests/e2e/online_serving/test_boogu_image.py
@@ -16,7 +16,7 @@ parity vs. the Diffusers baseline is covered separately (see the support plan, s
 From ``tests/``::
 
     pytest -s -v e2e/online_serving/test_boogu_image.py \
-        -m "full_model and diffusion and H100" --run-level=full_model
+        -m "slow and diffusion and H100"
 """
 
 import os
@@ -28,7 +28,7 @@ from tests.helpers.runtime import OmniServer, OmniServerParams, OpenAIClientHand
 
 os.environ["VLLM_WORKER_MULTIPROC_METHOD"] = "spawn"
 
-pytestmark = [pytest.mark.diffusion, pytest.mark.full_model]
+pytestmark = [pytest.mark.diffusion, pytest.mark.slow]
 
 MODEL = "Boogu/Boogu-Image-0.1-Base"
 T2I_PROMPT = "A mountain lake at sunset, photorealistic, cinematic lighting"
diff --git a/tests/e2e/online_serving/test_boogu_image_edit.py b/tests/e2e/online_serving/test_boogu_image_edit.py
index 45f999d26..d7b6d1dd1 100644
--- a/tests/e2e/online_serving/test_boogu_image_edit.py
+++ b/tests/e2e/online_serving/test_boogu_image_edit.py
@@ -19,7 +19,7 @@ reference image is supported.
 From ``tests/``::
 
     pytest -s -v e2e/online_serving/test_boogu_image_edit.py \
-        -m "full_model and diffusion and H100" --run-level=full_model
+        -m "slow and diffusion and H100"
 """
 
 import base64
@@ -35,7 +35,7 @@ from tests.helpers.runtime import (
     dummy_messages_from_mix_data,
 )
 
-pytestmark = [pytest.mark.diffusion, pytest.mark.full_model]
+pytestmark = [pytest.mark.diffusion, pytest.mark.slow]
 
 EDIT_PROMPT = "Change the style to a colored pencil drawing."
 SINGLE_CARD_FEATURE_MARKS = hardware_marks(res={"cuda": "H100"})
diff --git a/tests/e2e/online_serving/test_boogu_image_edit_expansion.py b/tests/e2e/online_serving/test_boogu_image_edit_expansion.py
index efc968eb6..1825aaafc 100644
--- a/tests/e2e/online_serving/test_boogu_image_edit_expansion.py
+++ b/tests/e2e/online_serving/test_boogu_image_edit_expansion.py
@@ -25,7 +25,7 @@ model and are intentionally omitted.
 From ``tests/``::
 
     pytest -s -v e2e/online_serving/test_boogu_image_edit_expansion.py \
-        -m "full_model and diffusion and H100" --run-level=full_model
+        -m "slow and diffusion and H100"
 """
 
 import os
@@ -38,7 +38,7 @@ from tests.helpers.runtime import OmniServer, OmniServerParams, OpenAIClientHand
 
 os.environ["VLLM_WORKER_MULTIPROC_METHOD"] = "spawn"
 
-pytestmark = [pytest.mark.diffusion, pytest.mark.full_model]
+pytestmark = [pytest.mark.diffusion, pytest.mark.slow]
 
 MODEL = "Boogu/Boogu-Image-0.1-Edit"
 EDIT_PROMPT = "Change the style to a colored pencil drawing."
diff --git a/tests/e2e/online_serving/test_boogu_image_expansion.py b/tests/e2e/online_serving/test_boogu_image_expansion.py
index ef6d3a62e..a56e5a338 100644
--- a/tests/e2e/online_serving/test_boogu_image_expansion.py
+++ b/tests/e2e/online_serving/test_boogu_image_expansion.py
@@ -30,7 +30,7 @@ omitted (add rows here when those features land).
 From ``tests/``::
 
     pytest -s -v e2e/online_serving/test_boogu_image_expansion.py \
-        -m "full_model and diffusion and H100" --run-level=full_model
+        -m "slow and diffusion and H100"
 """
 
 import os
@@ -42,7 +42,7 @@ from tests.helpers.runtime import OmniServer, OmniServerParams, OpenAIClientHand
 
 os.environ["VLLM_WORKER_MULTIPROC_METHOD"] = "spawn"
 
-pytestmark = [pytest.mark.diffusion, pytest.mark.full_model]
+pytestmark = [pytest.mark.diffusion, pytest.mark.slow]
 
 MODEL = "Boogu/Boogu-Image-0.1-Base"
 T2I_PROMPT = "A mountain lake at sunset, photorealistic, cinematic lighting"
diff --git a/tests/e2e/online_serving/test_glm_image_expansion.py b/tests/e2e/online_serving/test_glm_image_expansion.py
index d2a08f633..42e8d444a 100644
--- a/tests/e2e/online_serving/test_glm_image_expansion.py
+++ b/tests/e2e/online_serving/test_glm_image_expansion.py
@@ -30,7 +30,7 @@ from tests.helpers.runtime import (
 )
 from tests.helpers.stage_config import get_deploy_config_path, modify_stage_config
 
-pytestmark = [pytest.mark.diffusion, pytest.mark.full_model]
+pytestmark = [pytest.mark.diffusion, pytest.mark.slow]
 
 MODEL = os.environ.get("GLM_IMAGE_MODEL_PATH", "zai-org/GLM-Image")
 
diff --git a/tests/e2e/online_serving/test_lingbot_video_moe.py b/tests/e2e/online_serving/test_lingbot_video_moe.py
index 777880732..7e4d16568 100644
--- a/tests/e2e/online_serving/test_lingbot_video_moe.py
+++ b/tests/e2e/online_serving/test_lingbot_video_moe.py
@@ -63,7 +63,7 @@ def test_text_to_image_moe(omni_server: OmniServer, openai_client: OpenAIClientH
     )
 
 
-@pytest.mark.full_model
+@pytest.mark.slow
 @pytest.mark.diffusion
 @pytest.mark.parametrize("omni_server", _get_server_cases(MODEL), indirect=True)
 def test_video_generation_modes_moe(omni_server: OmniServer, openai_client: OpenAIClientHandler) -> None:
diff --git a/tests/e2e/online_serving/test_ltx.py b/tests/e2e/online_serving/test_ltx.py
index 43df5df2d..e4d8f51a8 100644
--- a/tests/e2e/online_serving/test_ltx.py
+++ b/tests/e2e/online_serving/test_ltx.py
@@ -10,7 +10,7 @@ from tests.helpers.mark import hardware_marks
 from tests.helpers.media import generate_synthetic_image
 from tests.helpers.runtime import OmniServer, OmniServerParams, OpenAIClientHandler
 
-pytestmark = [pytest.mark.diffusion, pytest.mark.full_model]
+pytestmark = [pytest.mark.diffusion, pytest.mark.slow]
 
 DISTILLED_MODEL = os.getenv("VLLM_TEST_LTX23_DISTILLED_MODEL", "diffusers/LTX-2.3-Distilled-Diffusers")
 BASE_MODEL = os.getenv("VLLM_TEST_LTX23_MODEL", "diffusers/LTX-2.3-Diffusers")
diff --git a/tests/e2e/online_serving/test_ltx25.py b/tests/e2e/online_serving/test_ltx25.py
index 2fad9f16a..da5c53851 100644
--- a/tests/e2e/online_serving/test_ltx25.py
+++ b/tests/e2e/online_serving/test_ltx25.py
@@ -20,7 +20,7 @@ MODEL = os.getenv("VLLM_TEST_LTX25_MODEL", DEFAULT_MODEL)
 MODEL_REVISION = os.getenv("VLLM_TEST_LTX25_MODEL_REVISION", DEFAULT_REVISION if MODEL == DEFAULT_MODEL else "")
 PROMPT = "A red fox walks through a snowy forest while the camera remains fixed."
 
-pytestmark = [pytest.mark.diffusion, pytest.mark.full_model]
+pytestmark = [pytest.mark.diffusion, pytest.mark.slow]
 SINGLE_CARD_MARKS = hardware_marks(res={"cuda": "H100"})
 
 
diff --git a/tests/e2e/online_serving/test_ming_flash_omni_expansion.py b/tests/e2e/online_serving/test_ming_flash_omni_expansion.py
index 8ab327022..6f6881d68 100644
--- a/tests/e2e/online_serving/test_ming_flash_omni_expansion.py
+++ b/tests/e2e/online_serving/test_ming_flash_omni_expansion.py
@@ -20,7 +20,7 @@ from tests.helpers.stage_config import get_deploy_config_path
 
 os.environ["VLLM_WORKER_MULTIPROC_METHOD"] = "spawn"
 
-pytestmark = [pytest.mark.omni, pytest.mark.full_model]
+pytestmark = [pytest.mark.omni, pytest.mark.slow]
 
 _SKIP_NEED_4_H100_NOT_CI = pytest.mark.skip(
     reason="Requires 4x H100 GPUs; skipped in CI for now.",
diff --git a/tests/e2e/online_serving/test_qwen_image_edit.py b/tests/e2e/online_serving/test_qwen_image_edit.py
index a02bc1b57..14d27318c 100644
--- a/tests/e2e/online_serving/test_qwen_image_edit.py
+++ b/tests/e2e/online_serving/test_qwen_image_edit.py
@@ -2,20 +2,19 @@
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 
 """
-Online serving tests for the Qwen-Image-Edit family (image-to-image via chat completions).
+Online serving tests for ``Qwen/Qwen-Image-Edit-2511`` (image-to-image via chat completions).
 
-- ``test_single_image_to_image_001``: ``Qwen/Qwen-Image-Edit`` — one reference image, fixed 512×512.
-- ``test_multi_images_to_image_001``: ``Qwen/Qwen-Image-Edit-2511`` — two reference images, fixed 512×512.
-- ``test_different_sizes_001``: ``Qwen/Qwen-Image-Edit-2511`` only, ``slow`` — mixed input
-  resolutions; ``extra_body`` uses per-output ``width``/``height`` lists; the test client sends one
-  scalar-size request per list index in parallel and merges images (see ``OpenAIClientHandler.send_diffusion_request``).
+- ``test_single_image_to_image_001``: one reference image, fixed 512×512.
+- ``test_multi_images_to_image_001``: two reference images, fixed 512×512.
+- ``test_different_sizes_001``: ``slow`` — mixed input resolutions; ``extra_body`` uses
+  per-output ``width``/``height`` lists; the test client sends one scalar-size request
+  per list index in parallel and merges images (see ``OpenAIClientHandler.send_diffusion_request``).
 
-``_get_diffusion_feature_cases`` registers a single ``default`` ``OmniServerParams`` row per model.
+``_get_diffusion_feature_cases`` registers a single ``default`` ``OmniServerParams`` row.
 
 From ``tests/``::
 
-    pytest -s -v e2e/online_serving/test_qwen_image_edit.py -m "core_model and diffusion" --run-level=core_model
-    pytest -s -v e2e/online_serving/test_qwen_image_edit.py -m "advanced_model and diffusion" --run-level=advanced_model
+    pytest -s -v e2e/online_serving/test_qwen_image_edit.py -m "slow and diffusion" --run-level=full_model
 """
 
 import pytest
@@ -29,6 +28,7 @@ from tests.helpers.runtime import (
     dummy_messages_from_mix_data,
 )
 
+MODEL = "Qwen/Qwen-Image-Edit-2511"
 EDIT_PROMPT = "Transform this modern, geometrist image into a Vincent van Gogh style impressionist painting."
 MULTI_EDIT_PROMPT = (
     "Transform the first image into a Dadaism collage art. "
@@ -39,7 +39,7 @@ NEGATIVE_PROMPT = "blurry, low quality, modern, geometrist"
 SINGLE_CARD_FEATURE_MARKS = hardware_marks(res={"cuda": "H100"})
 
 
-def _get_diffusion_feature_cases(model: str):
+def _get_diffusion_feature_cases(model: str = MODEL):
     """Return one ``default`` ``OmniServerParams`` row for ``model`` (no extra ``server_args``)."""
     return [
         pytest.param(
@@ -56,11 +56,11 @@ def _get_diffusion_feature_cases(model: str):
 @pytest.mark.diffusion
 @pytest.mark.parametrize(
     "omni_server",
-    _get_diffusion_feature_cases("Qwen/Qwen-Image-Edit"),
+    _get_diffusion_feature_cases(),
     indirect=True,
 )
 def test_single_image_to_image_001(omni_server: OmniServer, openai_client: OpenAIClientHandler):
-    """Single-reference edit smoke for ``Qwen/Qwen-Image-Edit`` (CFG when negative prompt + ``true_cfg_scale`` > 1)."""
+    """Single-reference edit smoke for ``Qwen/Qwen-Image-Edit-2511`` (CFG when negative prompt + ``true_cfg_scale`` > 1)."""
     image_data_url = f"data:image/jpeg;base64,{generate_synthetic_image(512, 512)['base64']}"
 
     messages = dummy_messages_from_mix_data(image_data_url=image_data_url, content_text=EDIT_PROMPT)
@@ -86,7 +86,7 @@ def test_single_image_to_image_001(omni_server: OmniServer, openai_client: OpenA
 @pytest.mark.diffusion
 @pytest.mark.parametrize(
     "omni_server",
-    _get_diffusion_feature_cases("Qwen/Qwen-Image-Edit-2511"),
+    _get_diffusion_feature_cases(),
     indirect=True,
 )
 def test_multi_images_to_image_001(omni_server: OmniServer, openai_client: OpenAIClientHandler):
@@ -117,7 +117,7 @@ def test_multi_images_to_image_001(omni_server: OmniServer, openai_client: OpenA
 @pytest.mark.diffusion
 @pytest.mark.parametrize(
     "omni_server",
-    _get_diffusion_feature_cases("Qwen/Qwen-Image-Edit-2511"),
+    _get_diffusion_feature_cases(),
     indirect=True,
 )
 def test_different_sizes_001(omni_server: OmniServer, openai_client: OpenAIClientHandler):
diff --git a/tests/e2e/online_serving/test_qwen_image_edit_expansion.py b/tests/e2e/online_serving/test_qwen_image_edit_expansion.py
index b3d6f8b19..e5e8c6c2b 100644
--- a/tests/e2e/online_serving/test_qwen_image_edit_expansion.py
+++ b/tests/e2e/online_serving/test_qwen_image_edit_expansion.py
@@ -1,8 +1,6 @@
 """
 Comprehensive tests of diffusion features that are available in online serving mode
-and are supported by the following models:
-- Qwen-Image-Edit: single image input
-- Qwen-Image-Edit-2511: single image input and two image inputs
+and are supported by Qwen-Image-Edit-2511 (single image input and two image inputs).
 """
 
 import pytest
@@ -11,10 +9,10 @@ from tests.helpers.mark import hardware_marks
 from tests.helpers.media import generate_synthetic_image
 from tests.helpers.runtime import OmniServer, OmniServerParams, OpenAIClientHandler, dummy_messages_from_mix_data
 
-pytestmark = [pytest.mark.diffusion, pytest.mark.full_model]
+pytestmark = [pytest.mark.diffusion, pytest.mark.slow]
 
-EDIT_PROMPT = "Transform this modern, geometrist image into a Vincent van Gogh style impressionist painting."
-SINGLE_EDIT_PROMPT_2511 = "Restyle this image into a Vincent van Gogh style impressionist painting."
+MODEL = "Qwen/Qwen-Image-Edit-2511"
+SINGLE_EDIT_PROMPT = "Restyle this image into a Vincent van Gogh style impressionist painting."
 MULTI_EDIT_PROMPT = (
     "Transform the first image into a Dadaism collage art. "
     "Transform the second image into a Vincent van Gogh style painting. "
@@ -25,9 +23,7 @@ SINGLE_CARD_FEATURE_MARKS = hardware_marks(res={"cuda": "H100"})
 PARALLEL_FEATURE_MARKS = hardware_marks(res={"cuda": "H100"}, num_cards=2)
 
 
-# This test file targets two models, so I write a helper function.
-# If a similar test only involves one model, one can just define a global list variable.
-def _get_diffusion_feature_cases(model: str):
+def _get_diffusion_feature_cases(model: str = MODEL):
     return [
         pytest.param(
             OmniServerParams(
@@ -123,35 +119,7 @@ def _get_diffusion_feature_cases(model: str):
 
 @pytest.mark.parametrize(
     "omni_server",
-    _get_diffusion_feature_cases("Qwen/Qwen-Image-Edit"),
-    indirect=True,
-)
-def test_qwen_image_edit(omni_server: OmniServer, openai_client: OpenAIClientHandler):
-    """Test all diffusion features with Qwen-Image-Edit in regular end-user scenarios."""
-    image_data_url = f"data:image/jpeg;base64,{generate_synthetic_image(512, 512)['base64']}"
-
-    messages = dummy_messages_from_mix_data(image_data_url=image_data_url, content_text=EDIT_PROMPT)
-
-    # CFG parallel is only activated when a negative prompt and true_cfg_scale > 1.0 are both present
-    request_config = {
-        "model": omni_server.model,
-        "messages": messages,
-        "extra_body": {
-            "height": 512,
-            "width": 512,
-            "num_inference_steps": 2,
-            "negative_prompt": NEGATIVE_PROMPT,
-            "true_cfg_scale": 4.0,
-            "seed": 42,
-        },
-    }
-
-    openai_client.send_diffusion_request(request_config)
-
-
-@pytest.mark.parametrize(
-    "omni_server",
-    _get_diffusion_feature_cases("Qwen/Qwen-Image-Edit-2511"),
+    _get_diffusion_feature_cases(),
     indirect=True,
 )
 def test_qwen_image_edit_2511_single_image(omni_server: OmniServer, openai_client: OpenAIClientHandler):
@@ -166,7 +134,7 @@ def test_qwen_image_edit_2511_single_image(omni_server: OmniServer, openai_clien
     """
     image_data_url = f"data:image/jpeg;base64,{generate_synthetic_image(512, 512)['base64']}"
 
-    messages = dummy_messages_from_mix_data(image_data_url=image_data_url, content_text=SINGLE_EDIT_PROMPT_2511)
+    messages = dummy_messages_from_mix_data(image_data_url=image_data_url, content_text=SINGLE_EDIT_PROMPT)
 
     request_config = {
         "model": omni_server.model,
@@ -186,7 +154,7 @@ def test_qwen_image_edit_2511_single_image(omni_server: OmniServer, openai_clien
 
 @pytest.mark.parametrize(
     "omni_server",
-    _get_diffusion_feature_cases("Qwen/Qwen-Image-Edit-2511"),
+    _get_diffusion_feature_cases(),
     indirect=True,
 )
 def test_qwen_image_edit_2511_two_images(omni_server: OmniServer, openai_client: OpenAIClientHandler):
diff --git a/tests/e2e/online_serving/test_qwen_image_layered_expansion.py b/tests/e2e/online_serving/test_qwen_image_layered_expansion.py
index 315fba6ef..8c5001a4a 100644
--- a/tests/e2e/online_serving/test_qwen_image_layered_expansion.py
+++ b/tests/e2e/online_serving/test_qwen_image_layered_expansion.py
@@ -18,7 +18,7 @@ from tests.helpers.mark import hardware_marks
 from tests.helpers.media import decode_b64_image, generate_synthetic_image
 from tests.helpers.runtime import OmniServer, OmniServerParams, OpenAIClientHandler, dummy_messages_from_mix_data
 
-pytestmark = [pytest.mark.diffusion, pytest.mark.full_model]
+pytestmark = [pytest.mark.diffusion, pytest.mark.slow]
 
 MODEL = "Qwen/Qwen-Image-Layered"
 EDIT_PROMPT = "Decompose this image into layers."
diff --git a/tests/e2e/online_serving/test_voxcpm2_tts_expansion.py b/tests/e2e/online_serving/test_voxcpm2_tts_expansion.py
index 996585a1d..de73ae6da 100644
--- a/tests/e2e/online_serving/test_voxcpm2_tts_expansion.py
+++ b/tests/e2e/online_serving/test_voxcpm2_tts_expansion.py
@@ -16,7 +16,7 @@ from tests.helpers.media import load_test_audio_data_url
 from tests.helpers.runtime import OmniServerParams
 from tests.helpers.stage_config import get_deploy_config_path
 
-pytestmark = [pytest.mark.full_model, pytest.mark.tts]
+pytestmark = [pytest.mark.slow, pytest.mark.tts]
 
 os.environ["VLLM_WORKER_MULTIPROC_METHOD"] = "spawn"
 
diff --git a/tests/e2e/online_serving/test_wan_2_1_vace_expansion.py b/tests/e2e/online_serving/test_wan_2_1_vace_expansion.py
index d58bb9e32..4c6380314 100644
--- a/tests/e2e/online_serving/test_wan_2_1_vace_expansion.py
+++ b/tests/e2e/online_serving/test_wan_2_1_vace_expansion.py
@@ -26,7 +26,7 @@ import pytest
 from tests.helpers.mark import hardware_marks
 from tests.helpers.runtime import OmniServer, OmniServerParams, OpenAIClientHandler
 
-pytestmark = [pytest.mark.diffusion, pytest.mark.full_model]
+pytestmark = [pytest.mark.diffusion, pytest.mark.slow]
 
 MODEL = "Wan-AI/Wan2.1-VACE-1.3B-diffusers"
 PROMPT = "A cat walking slowly across a sunlit garden path"
diff --git a/tools/nightly/buildkite_testcase_statistics.py b/tools/nightly/buildkite_testcase_statistics.py
index 740f896f5..06f09c791 100644
--- a/tools/nightly/buildkite_testcase_statistics.py
+++ b/tools/nightly/buildkite_testcase_statistics.py
@@ -1,8 +1,7 @@
 #!/usr/bin/env python3
 """
-Parse pytest commands from Buildkite test-ready.yml, test-merge.yml, test-nightly.yml,
-and test-weekly.yml; collect test cases (including parametrized) via pytest --collect-only -q
-and produce an HTML report.
+Parse pytest commands from Buildkite .buildkite/cuda/test-{ready,merge,nightly,weekly}.yml;
+collect test cases (including parametrized) via pytest --collect-only -q and produce an HTML report.
 
 Leaf steps may live under ``group:`` blocks (nested ``steps``); those are flattened with the
 group title carried into the report.
@@ -30,7 +29,7 @@ import yaml
 
 # Repo root (parent of the directory containing this script)
 REPO_ROOT = Path(__file__).resolve().parent.parent.parent
-BUILDKITE_DIR = REPO_ROOT / ".buildkite"
+BUILDKITE_DIR = REPO_ROOT / ".buildkite" / "cuda"
 PIPELINE_FILES = ["test-ready.yml", "test-merge.yml", "test-nightly.yml", "test-weekly.yml"]
 
 
@@ -256,13 +255,34 @@ def _resolve_pytest_target(target: str, repo_root: Path, extra: list[str], raw_l
         rel_dir = Path(target.rstrip("/"))
         dir_path = (repo_root / rel_dir).resolve()
         if dir_path.exists() and dir_path.is_dir():
-            return [str(dir_path)], target.replace("\\", "/").rstrip("/"), 120
+            return [str(dir_path)], target.replace("\\", "/").rstrip("/"), 300
 
     if not extra:
         raise RuntimeError(f"Failed to parse pytest sidecar args from line: {raw_line!r}")
     if not (repo_root / "tests").exists():
         raise FileNotFoundError("tests/ directory not found under repo root")
-    return ["tests/"], "tests/", 120
+    # Broad ``tests/`` collection loads many conftests; allow more time on busy CI agents.
+    return ["tests/"], "tests/", 300
+
+
+def _format_pytest_collect_failure(
+    path_args: list[str],
+    extra: list[str],
+    result: subprocess.CompletedProcess[str],
+) -> str:
+    """pytest often puts collection errors on stdout; include both streams + exit code."""
+    stdout = (result.stdout or "").strip()
+    stderr = (result.stderr or "").strip()
+    parts = [
+        f"pytest --collect-only failed for {path_args} with args {extra} (exit={result.returncode})",
+    ]
+    if stderr:
+        parts.append(f"stderr:\n{stderr}")
+    if stdout:
+        parts.append(f"stdout:\n{stdout}")
+    if not stderr and not stdout:
+        parts.append("(no stdout/stderr captured)")
+    return "\n".join(parts)
 
 
 def collect_test_names(target: str, repo_root: Path, raw_line: str) -> list[str]:
@@ -273,6 +293,8 @@ def collect_test_names(target: str, repo_root: Path, raw_line: str) -> list[str]
     flags ``--ignore``, ``--test-config-file``, ``-m``, ``-k``, ``--run-level`` are taken from the
     full ``raw_line`` and appended to the same invocation so collection matches CI (perf scripts
     parametrize from ``--test-config-file`` at import time).
+
+    Pytest exit code 5 (no tests collected) is treated as an empty result, not a hard failure.
     """
     extra = _pytest_collect_sidecar_args(raw_line)
     path_args, _fallback, timeout_quiet = _resolve_pytest_target(target, repo_root, extra, raw_line)
@@ -286,8 +308,15 @@ def collect_test_names(target: str, repo_root: Path, raw_line: str) -> list[str]
         text=True,
         timeout=timeout_quiet,
     )
+    # 5 == no tests collected under the given -m/-k filters.
+    if result.returncode == 5:
+        print(
+            f"pytest --collect-only: no tests matched for {path_args} with args {extra}",
+            file=sys.stderr,
+        )
+        return []
     if result.returncode != 0:
-        raise RuntimeError(f"pytest --collect-only failed for {path_args} with args {extra}: {result.stderr.strip()}")
+        raise RuntimeError(_format_pytest_collect_failure(path_args, extra, result))
     print(f"pytest --collect-only success for {path_args} with args {extra}: {result.stdout.strip()}")
     return _parse_collect_only_stdout(result.stdout or "", stderr=result.stderr or "")
 
@@ -962,7 +991,7 @@ def main() -> None:
         "--buildkite-dir",
         type=Path,
         default=BUILDKITE_DIR,
-        help="Path to .buildkite directory",
+        help="Path to Buildkite CUDA pipeline directory (default: .buildkite/cuda)",
     )
     default_out = REPO_ROOT / "buildkite_testcase_statistics.html"
     parser.add_argument(
@@ -1023,7 +1052,12 @@ def main() -> None:
             names = []
             skipped_ids: set[str] = set()
             for target, raw in targets:
-                collected = collect_test_names(target, REPO_ROOT, raw)
+                try:
+                    collected = collect_test_names(target, REPO_ROOT, raw)
+                except (RuntimeError, FileNotFoundError, subprocess.TimeoutExpired) as exc:
+                    # Keep generating the report for other steps; surface the root cause.
+                    print(f"WARNING: collect failed for {module_key} target={target!r}: {exc}", file=sys.stderr)
+                    collected = []
                 skipped_ids.update(get_skip_status(target, raw, REPO_ROOT))
                 names.extend(collected)
             count = len(names) - len(skipped_ids)