diff --git a/benchmarks/multi_node/srt-slurm-recipes/sglang/deepseek-v4/agentic/agg-b300-dep4-mtp-kvoffload.yaml b/benchmarks/multi_node/srt-slurm-recipes/sglang/deepseek-v4/agentic/agg-b300-dep4-mtp-kvoffload.yaml
new file mode 100644
index 000000000..fe5ba9d5a
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/sglang/deepseek-v4/agentic/agg-b300-dep4-mtp-kvoffload.yaml
@@ -0,0 +1,137 @@
+name: "agg-b300-dep4-mtp-kvoffload"
+
+model:
+  path: "deepseek-v4-pro-0813"
+  container: "dynamo-sglang"
+  precision: "fp4"
+
+identity:
+  model:
+    repo: "deepseek-ai/DeepSeek-V4-Pro-0813"
+  container:
+    image: "lmsysorg/sglang:nightly-dev-cu13-20260816-4a6dc267"
+
+dynamo:
+  install: true
+  hash: "7a4e47ead90ec6b479e397996f6479ce7d251510"
+
+slurm:
+  time_limit: "8:00:00"
+
+health_check:
+  max_attempts: 1440
+  interval_seconds: 10
+
+resources:
+  gpu_type: "b300"
+  gpus_per_node: 4
+  agg_nodes: 1
+  agg_workers: 1
+  gpus_per_agg: 4
+
+infra:
+  etcd_nats_dedicated_node: true
+  nats_max_payload_mb: 32
+
+frontend:
+  type: dynamo
+  nginx_session_affinity: true
+  nginx_session_affinity_header: X-Dynamo-Session-ID
+  enable_multiple_frontends: false
+  env:
+    PIP_BREAK_SYSTEM_PACKAGES: "1"
+  args:
+    router-mode: "kv"
+    router-session-affinity-ttl-secs: "14400"
+    active-decode-blocks-threshold: "None"
+    active-prefill-tokens-threshold: "None"
+    active-prefill-tokens-threshold-frac: "None"
+
+backend:
+  type: sglang
+  # Let srt-slurm allocate and register the worker-local publisher endpoint;
+  # Dynamo needs discovery metadata in addition to the SGLang CLI flag.
+  kv_events_config:
+    aggregated: true
+  aggregated_environment:
+    PIP_BREAK_SYSTEM_PACKAGES: "1"
+    SGLANG_RAGGED_VERIFY_MODE: "static"
+    SGLANG_DEFAULT_THINKING: "1"
+    SGLANG_DSV4_REASONING_EFFORT: high
+    SGLANG_ENABLE_UNIFIED_RADIX_TREE: "1"
+    SGLANG_OPT_UNIFIED_CACHE_FREE_OUT_OF_WINDOW_SLOTS: "1"
+    SGLANG_OPT_USE_DEEPGEMM_MEGA_MOE: "1"
+    SGLANG_OPT_DEEPGEMM_MEGA_MOE_USE_FP4_ACTS: "1"
+    SGLANG_OPT_DEEPGEMM_MEGA_MOE_USE_MXF4_KIND: "1"
+    SGLANG_OPT_DEEPGEMM_MEGA_MOE_NUM_MAX_TOKENS_PER_RANK: "9216"
+    SGLANG_OPT_USE_CUSTOM_ALL_REDUCE_V2: "0"
+    SGLANG_OPT_SWA_SPLIT_LEAF_ON_INSERT: "1"
+    SGLANG_OPT_USE_JIT_INDEXER_METADATA: "1"
+    SGLANG_OPT_USE_JIT_NORM: "1"
+    SGLANG_OPT_USE_TOPK_V2: "1"
+  sglang_config:
+    aggregated:
+      served-model-name: "deepseek-ai/DeepSeek-V4-Pro-0813"
+      trust-remote-code: true
+      enable-metrics: true
+      enable-cache-report: true
+      # The AgentX corpus contains a small number of traces longer than the
+      # runtime KV capacity. Match the existing native B300 launcher: truncate
+      # those requests instead of leaving Dynamo streams unresolved.
+      allow-auto-truncate: true
+      # DSpark draft weights raise the measured minimum viable fraction to
+      # 0.9258 on B300; use the same 0.94 fraction as the GB300 TP4 recipe.
+      mem-fraction-static: 0.94
+      swa-full-tokens-ratio: 0.02
+      chunked-prefill-size: 32768
+      max-prefill-tokens: 8192
+      max-running-requests: 128
+      cuda-graph-max-bs: 32
+      stream-interval: 20
+      incremental-streaming-output: true
+      watchdog-timeout: 1800
+      dp-size: 4
+      tp-size: 4
+      ep-size: 4
+      enable-dp-attention: true
+      enable-dp-lm-head: true
+      enable-prefill-delayer: true
+      prefill-delayer-max-delay-passes: 30
+      moe-dense-tp-size: 1
+      load-balance-method: total_tokens
+      moe-a2a-backend: megamoe
+      enable-hierarchical-cache: true
+      # DEP4's smaller device KV pool otherwise leaves only ~28 GB of host
+      # prefix cache per node; ratio 75 raises it to ~700 GB, matching the
+      # effective per-session capacity of the validated DEP8 configuration.
+      hicache-ratio: 75
+      hicache-write-policy: write_back
+      hicache-io-backend: direct
+      hicache-mem-layout: page_first_direct
+      weight-loader-drop-cache-after-load: true
+      speculative-algorithm: DSPARK
+      speculative-dspark-block-size: 6
+      speculative-num-steps: 1
+      speculative-eagle-topk: 1
+      speculative-num-draft-tokens: 7
+
+sbatch_directives:
+  mem: "0"
+  cpus-per-task: "144"
+
+srun_options:
+  mem: "0"
+  container-remap-root: ""
+
+benchmark:
+  type: custom
+  command: bash /infmax-workspace/benchmarks/multi_node/agentic_srt.sh
+  env:
+    INFMAX_CONTAINER_WORKSPACE: /infmax-workspace
+    RESULT_DIR: /logs/agentic
+    PORT: "8000"
+    IS_MULTINODE: "false"
+    TP: "4"
+    AIPERF_HTTP_X_DYNAMO_SESSION_ID_FROM_CORRELATION_ID: "true"
+    AIPERF_USE_DYNAMO_CONV_AWARE_ROUTING: "0"
+    AIPERF_REQUIRED_SERVER_METRIC_PREFIX: "sglang:"
diff --git a/benchmarks/multi_node/srt-slurm-recipes/sglang/deepseek-v4/agentic/agg-b300-dep8-mtp-kvoffload.yaml b/benchmarks/multi_node/srt-slurm-recipes/sglang/deepseek-v4/agentic/agg-b300-dep8-mtp-kvoffload.yaml
new file mode 100644
index 000000000..a57e8c367
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/sglang/deepseek-v4/agentic/agg-b300-dep8-mtp-kvoffload.yaml
@@ -0,0 +1,136 @@
+name: "agg-b300-dep8-mtp-kvoffload"
+
+model:
+  path: "deepseek-v4-pro-0813"
+  container: "dynamo-sglang"
+  precision: "fp4"
+
+identity:
+  model:
+    repo: "deepseek-ai/DeepSeek-V4-Pro-0813"
+  container:
+    image: "lmsysorg/sglang:nightly-dev-cu13-20260816-4a6dc267"
+
+dynamo:
+  install: true
+  hash: "7a4e47ead90ec6b479e397996f6479ce7d251510"
+
+slurm:
+  time_limit: "8:00:00"
+
+health_check:
+  max_attempts: 1440
+  interval_seconds: 10
+
+resources:
+  gpu_type: "b300"
+  gpus_per_node: 8
+  agg_nodes: 1
+  agg_workers: 1
+  gpus_per_agg: 8
+
+infra:
+  etcd_nats_dedicated_node: true
+  nats_max_payload_mb: 32
+
+frontend:
+  type: dynamo
+  nginx_session_affinity: true
+  nginx_session_affinity_header: X-Dynamo-Session-ID
+  enable_multiple_frontends: false
+  env:
+    PIP_BREAK_SYSTEM_PACKAGES: "1"
+  args:
+    router-mode: "kv"
+    router-session-affinity-ttl-secs: "14400"
+    active-decode-blocks-threshold: "None"
+    active-prefill-tokens-threshold: "None"
+    active-prefill-tokens-threshold-frac: "None"
+
+backend:
+  type: sglang
+  # Let srt-slurm allocate and register the worker-local publisher endpoint;
+  # Dynamo needs discovery metadata in addition to the SGLang CLI flag.
+  kv_events_config:
+    aggregated: true
+  aggregated_environment:
+    PIP_BREAK_SYSTEM_PACKAGES: "1"
+    SGLANG_RAGGED_VERIFY_MODE: "static"
+    SGLANG_DEFAULT_THINKING: "1"
+    SGLANG_DSV4_REASONING_EFFORT: high
+    SGLANG_ENABLE_UNIFIED_RADIX_TREE: "1"
+    SGLANG_OPT_UNIFIED_CACHE_FREE_OUT_OF_WINDOW_SLOTS: "1"
+    SGLANG_OPT_USE_DEEPGEMM_MEGA_MOE: "1"
+    SGLANG_OPT_DEEPGEMM_MEGA_MOE_USE_FP4_ACTS: "1"
+    SGLANG_OPT_DEEPGEMM_MEGA_MOE_USE_MXF4_KIND: "1"
+    SGLANG_OPT_DEEPGEMM_MEGA_MOE_NUM_MAX_TOKENS_PER_RANK: "9216"
+    SGLANG_OPT_USE_CUSTOM_ALL_REDUCE_V2: "0"
+    SGLANG_OPT_SWA_SPLIT_LEAF_ON_INSERT: "1"
+    SGLANG_OPT_USE_JIT_INDEXER_METADATA: "1"
+    SGLANG_OPT_USE_JIT_NORM: "1"
+    SGLANG_OPT_USE_TOPK_V2: "1"
+  sglang_config:
+    aggregated:
+      served-model-name: "deepseek-ai/DeepSeek-V4-Pro-0813"
+      trust-remote-code: true
+      enable-metrics: true
+      enable-cache-report: true
+      # The AgentX corpus contains a small number of traces longer than the
+      # runtime KV capacity. Match the existing native B300 launcher: truncate
+      # those requests instead of leaving Dynamo streams unresolved.
+      allow-auto-truncate: true
+      # Leave activation headroom for 8K/rank chunked prefill; 0.90 OOMs
+      # under the c128 AgentX warmup while 0.85 completes the full run.
+      mem-fraction-static: 0.85
+      swa-full-tokens-ratio: 0.02
+      chunked-prefill-size: 65536
+      max-prefill-tokens: 8192
+      max-running-requests: 1152
+      cuda-graph-max-bs: 144
+      stream-interval: 20
+      incremental-streaming-output: true
+      watchdog-timeout: 1800
+      dp-size: 8
+      tp-size: 8
+      ep-size: 8
+      enable-dp-attention: true
+      enable-dp-lm-head: true
+      enable-prefill-delayer: true
+      prefill-delayer-max-delay-passes: 30
+      moe-dense-tp-size: 1
+      load-balance-method: total_tokens
+      moe-a2a-backend: megamoe
+      enable-hierarchical-cache: true
+      # Keep the explicit host KV pool within the x86 B300 node's 3 TB DRAM
+      # budget; ratio 2.0 OOMs during startup before the benchmark can run.
+      hicache-ratio: 1.5
+      hicache-write-policy: write_back
+      hicache-io-backend: direct
+      hicache-mem-layout: page_first_direct
+      weight-loader-drop-cache-after-load: true
+      speculative-algorithm: DSPARK
+      speculative-dspark-block-size: 6
+      speculative-num-steps: 1
+      speculative-eagle-topk: 1
+      speculative-num-draft-tokens: 7
+
+sbatch_directives:
+  mem: "0"
+  cpus-per-task: "144"
+
+srun_options:
+  mem: "0"
+  container-remap-root: ""
+
+benchmark:
+  type: custom
+  command: bash /infmax-workspace/benchmarks/multi_node/agentic_srt.sh
+  env:
+    INFMAX_CONTAINER_WORKSPACE: /infmax-workspace
+    RESULT_DIR: /logs/agentic
+    PORT: "8000"
+    IS_MULTINODE: "false"
+    TP: "8"
+    AIPERF_HTTP_X_DYNAMO_SESSION_ID_FROM_CORRELATION_ID: "true"
+    AIPERF_USE_DYNAMO_CONV_AWARE_ROUTING: "0"
+    AIPERF_REQUIRED_SERVER_METRIC_PREFIX: "sglang:"
diff --git a/configs/nvidia-master.yaml b/configs/nvidia-master.yaml
index 2f73b1c64..ac7b96152 100644
--- a/configs/nvidia-master.yaml
+++ b/configs/nvidia-master.yaml
@@ -8705,6 +8705,7 @@ glm5.1-fp8-b200-tilert:
           dp-attn: false
           additional-settings:
           - "DECODE_NODES=1"
+
     - isl: 8192
       osl: 1024
       search-space:
@@ -8725,3 +8726,54 @@ glm5.1-fp8-b200-tilert:
           dp-attn: false
           additional-settings:
           - "DECODE_NODES=1"
+
+dsv4-fp4-b300-dynamo-sglang-agentic-mtp-agg:
+  image: lmsysorg/sglang:nightly-dev-cu13-20260816-4a6dc267
+  model: deepseek-ai/DeepSeek-V4-Pro-0813
+  model-prefix: dsv4
+  runner: cluster:b300-nv
+  precision: fp4
+  framework: dynamo-sglang
+  router: { name: dynamo-router, version: "7a4e47ead90ec6b479e397996f6479ce7d251510" }
+  multinode: true
+  disagg: false
+  scenarios:
+    agentic-coding:
+    - dram-utilization: 0.95
+      search-space:
+      - spec-decoding: draft_model
+        conc-list: [48, 64]
+        kv-offloading: dram
+        kv-offload-backend: { name: hicache }
+        prefill:
+          num-worker: 1
+          tp: 4
+          ep: 4
+          dp-attn: true
+          additional-settings:
+          - "SYNTHETIC_ACCEPTANCE=true"
+          - "SYNTHETIC_ACCEPTANCE_LENGTH=3.77"
+          - "CONFIG_FILE=recipes/sglang/deepseek-v4/agentic/agg-b300-dep4-mtp-kvoffload.yaml"
+        decode:
+          num-worker: 0
+          tp: 4
+          ep: 4
+          dp-attn: true
+      - spec-decoding: draft_model
+        conc-list: [128, 256, 384, 512, 576]
+        kv-offloading: dram
+        kv-offload-backend: { name: hicache }
+        prefill:
+          num-worker: 1
+          tp: 8
+          ep: 8
+          dp-attn: true
+          additional-settings:
+          - "SYNTHETIC_ACCEPTANCE=true"
+          - "SYNTHETIC_ACCEPTANCE_LENGTH=3.77"
+          - "CONFIG_FILE=recipes/sglang/deepseek-v4/agentic/agg-b300-dep8-mtp-kvoffload.yaml"
+        decode:
+          num-worker: 0
+          tp: 8
+          ep: 8
+          dp-attn: true
diff --git a/golden_al_distribution/dsv4_dspark_probabilistic.yaml b/golden_al_distribution/dsv4_dspark_probabilistic.yaml
new file mode 100644
index 000000000..f72fe9431
--- /dev/null
+++ b/golden_al_distribution/dsv4_dspark_probabilistic.yaml
@@ -0,0 +1,19 @@
+deepseek-v4-pro-0813:
+  thinking_on:
+    1: 1.84
+    2: 2.51
+    3: 3.01
+    4: 3.36
+    5: 3.61
+    6: 3.77
+    7: 3.73
+    8: 3.47
+  thinking_off:
+    1: 1.94
+    2: 2.81
+    3: 3.58
+    4: 4.17
+    5: 4.74
+    6: 5.12
+    7: 5.24
+    8: 4.95
diff --git a/perf-changelog.yaml b/perf-changelog.yaml
index 085e9cb88..f1b0b98d9 100644
--- a/perf-changelog.yaml
+++ b/perf-changelog.yaml
@@ -6039,7 +6039,6 @@
     - "EVAL_ONLY runs switch to real block verification by removing synthetic acceptance (TLLM_SPEC_DECODE_FORCE_NUM_ACCEPTED_TOKENS stripped from all agentx recipes before srtctl)."
     - "Add SERVED_MODEL_NAME: Qwen3.5-397B-A17B-NVFP4-V2 to all 6 agentx recipe env blocks and make build_replay_cmd prefer SERVED_MODEL_NAME over $MODEL for the aiperf --model flag."
   pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/2612
-
 - config-keys:
     - glm5.2-fp4-gb200-dynamo-sglang-agentic-agg
     - glm5.2-fp4-gb200-dynamo-sglang-agentic-disagg
@@ -6049,3 +6048,14 @@
     - "Add GB200 GLM-5.2 FP4 Dynamo-SGLang AgentX with EAGLE MTP, HiCache DRAM offload, and KV-aware correlation-ID affinity."
     - "Measure the TP8 aggregate curve and tuned 1P1D TP4 HiCache c10/c12 disaggregated frontier with every logical SGLang metrics endpoint."
   pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/2620
+
+- config-keys:
+    - dsv4-fp4-b300-dynamo-sglang-agentic-mtp-agg
+  scenario-type:
+    - agentic-coding
+  description:
+    - "Add B300 Dynamo-SGLang DSpark6 AgentX aggregate DEP4 and DEP8 configurations with session-affinity routing, HiCache host offload, an 8K prefill window, and thinking-on block-size-6 golden synthetic acceptance length 3.77 for throughput; eval retains real verification."
+    - "Use the August 16 SGLang CUDA 13 nightly and srt-slurm v1.0.38, with DEP4 concurrency 48-64 and DEP8 concurrency 128-576."
+    - "Cap the DEP8 HiCache ratio at 1.5 so the explicit host KV pool fits within the B300 x86 node's 3 TB DRAM budget during model loading and JIT startup."
+    - "Use the official DeepSeek-V4-Pro-0813 model identity so AIPerf follows the standard --tokenizer $MODEL path and reuses the shared Hugging Face cache."
+  pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/2624
diff --git a/runners/launch_b300-nv.sh b/runners/launch_b300-nv.sh
index f7878a8ad..d42ef5e66 100644
--- a/runners/launch_b300-nv.sh
+++ b/runners/launch_b300-nv.sh
@@ -1,5 +1,7 @@
 #!/usr/bin/bash
 
+source "$(dirname "${BASH_SOURCE[0]}")/slurm_utils.sh"
+
 # System-specific configuration for B300 NV Slurm cluster (sa-shared)
 SLURM_PARTITION="batch_1"
 SLURM_ACCOUNT="benchmark"
@@ -41,6 +43,9 @@ elif [[ $MODEL_PREFIX == "dsv4" && $PRECISION == "fp4" && $FRAMEWORK == "dynamo-
     fi
     export MODEL_PATH="${SELECTED_MODEL_PATH:-/data/models/dsv4-pro}"
     export SRT_SLURM_MODEL_PREFIX="deepseek-v4-pro"
+elif [[ $MODEL_PREFIX == "dsv4" && $PRECISION == "fp4" && $FRAMEWORK == "dynamo-sglang" && $MODEL == "deepseek-ai/DeepSeek-V4-Pro-0813" ]]; then
+    export MODEL_PATH="${MODEL_PATH:-/scratch/models/DeepSeek-V4-Pro-0813}"
+    export SRT_SLURM_MODEL_PREFIX="deepseek-v4-pro-0813"
 elif [[ $MODEL_PREFIX == "dsv4" && $PRECISION == "fp4" && $FRAMEWORK == "dynamo-sglang" ]]; then
     export MODEL_PATH="${MODEL_PATH:-/scratch/models/DeepSeek-V4-Pro}"
     export SRT_SLURM_MODEL_PREFIX="deepseek-v4-pro"
@@ -70,7 +75,17 @@ if [ -d "$SRT_REPO_DIR" ]; then
 fi
 
 # TODO(CJQ): make first class upon srt-slurm upstream refactor
-if [[ "$IS_AGENTIC" == "1" ]]; then
+if [[ "$IS_AGENTIC" == "1" && $FRAMEWORK == "dynamo-sglang" && $MODEL_PREFIX == "dsv4" && $PRECISION == "fp4" ]]; then
+    # Match the GB300 DSV4 AgentX path: v1.0.38 provides --no-preflight,
+    # session-affinity frontend support, backend metric injection for custom
+    # benchmarks, and the current aggregated Dynamo-SGLang recipe schema.
+    git clone --branch v1.0.38 --single-branch https://github.com/NVIDIA/srt-slurm.git "$SRT_REPO_DIR" || exit 1
+    cd "$SRT_REPO_DIR" || exit 1
+    mkdir -p recipes/sglang/deepseek-v4/agentic
+    cp -R \
+        "$GITHUB_WORKSPACE/benchmarks/multi_node/srt-slurm-recipes/sglang/deepseek-v4/agentic/." \
+        recipes/sglang/deepseek-v4/agentic
+elif [[ "$IS_AGENTIC" == "1" ]]; then
     git clone --branch cam/sa-submission-q2-2026 --single-branch https://github.com/cquil11/srt-slurm-nv.git "$SRT_REPO_DIR"
     cd "$SRT_REPO_DIR" || exit 1
 elif [[ $FRAMEWORK == "dynamo-vllm" && $MODEL_PREFIX == "dsv4" ]]; then
@@ -135,9 +150,45 @@ NGINX_IMAGE="nginx:1.27.4"
 SQUASH_FILE="/data/squash/$(echo "$IMAGE" | sed 's/[\/:@#]/_/g').sqsh"
 NGINX_SQUASH_FILE="/data/squash/$(echo "$NGINX_IMAGE" | sed 's/[\/:@#]/_/g').sqsh"
 
-# Import containers via enroot
-srun -N 1 -A $SLURM_ACCOUNT -p $SLURM_PARTITION bash -c "enroot import -o $SQUASH_FILE docker://$IMAGE"
-srun -N 1 -A $SLURM_ACCOUNT -p $SLURM_PARTITION bash -c "enroot import -o $NGINX_SQUASH_FILE docker://$NGINX_IMAGE"
+# Import containers via enroot. Perform the validity check on the allocated
+# compute node, not before submitting srun: a job can wait in the queue while
+# another runner creates the same shared squash file. Serialize imports so a
+# concurrent cache miss cannot race on the output either.
+import_squash() {
+    local squash_file="$1"
+    local image_ref="$2"
+
+    srun -N 1 -A "$SLURM_ACCOUNT" -p "$SLURM_PARTITION" \
+        bash -s -- "$squash_file" "$image_ref" <<'IMPORT_SQUASH'
+set -euo pipefail
+squash_file="$1"
+image_ref="$2"
+lock_dir="$(dirname "$squash_file")/.locks"
+mkdir -p "$lock_dir"
+lock_file="$lock_dir/$(basename "$squash_file").lock"
+
+(
+    flock -w 3600 9 || {
+        echo "Failed to acquire lock for $squash_file" >&2
+        exit 1
+    }
+    if unsquashfs -l "$squash_file" > /dev/null 2>&1; then
+        echo "Squash file already exists and is valid, skipping import: $squash_file"
+    else
+        rm -f "$squash_file"
+        enroot import -o "$squash_file" "docker://$image_ref"
+        unsquashfs -l "$squash_file" > /dev/null 2>&1 || {
+            echo "enroot import did not produce a valid squash file: $squash_file" >&2
+            exit 1
+        }
+        chmod a+r "$squash_file" || true
+    fi
+) 9>"$lock_file"
+IMPORT_SQUASH
+}
+
+import_squash "$SQUASH_FILE" "$IMAGE"
+import_squash "$NGINX_SQUASH_FILE" "$NGINX_IMAGE"
 
 export ISL="$ISL"
 export OSL="$OSL"
@@ -171,6 +222,7 @@ containers:
 use_exclusive_sbatch_directive: true
 default_mounts:
   "/opt/ucx-no-ud": "/usr/local/ucx"
+  "${GITHUB_WORKSPACE}": "/infmax-workspace"
 EOF
 
 echo "Generated srtslurm.yaml:"
@@ -200,6 +252,11 @@ fi
 
 # Override the job name in the recipe with the runner name.
 sed -i "s/^name:.*/name: \"${RUNNER_NAME}\"/" "$CONFIG_PATH"
+
+# Throughput recipes opt into synthetic acceptance through the master config.
+# Eval-only jobs leave the checked-in real-MTP recipe unchanged so generated
+# tokens still pass target-model verification.
+inject_synthetic_acceptance "$CONFIG_PATH" "$FRAMEWORK" || exit 1
 if [[ "$MODEL_PREFIX" == "minimaxm3" && -n "$MINIMAX_M3_SLURM_EXCLUDED_NODELIST" ]]; then
     sed -i "/^name:.*/a sbatch_directives:\n  exclude: \"${MINIMAX_M3_SLURM_EXCLUDED_NODELIST}\"" "$CONFIG_PATH"
 fi
@@ -483,9 +540,17 @@ else
     export GPU_COUNT="${GPU_COUNT:-${TP:?TP must be set}}"
 
     SALLOC_TIME_LIMIT="${SALLOC_TIME_LIMIT:-480}"
-    salloc --partition=$SLURM_PARTITION --account=$SLURM_ACCOUNT -N 1 --gres=gpu:$GPU_COUNT --exclusive --mem=0 --time="$SALLOC_TIME_LIMIT" --no-shell --job-name="$RUNNER_NAME"
+    if ! salloc --partition=$SLURM_PARTITION --account=$SLURM_ACCOUNT -N 1 --gres=gpu:$GPU_COUNT --exclusive --mem=0 --time="$SALLOC_TIME_LIMIT" --no-shell --job-name="$RUNNER_NAME"; then
+        echo "Failed to allocate B300 Slurm resources for account=$SLURM_ACCOUNT partition=$SLURM_PARTITION" >&2
+        exit 1
+    fi
     JOB_ID=$(squeue --name="$RUNNER_NAME" -u "$USER" -h -o %A | head -n1)
 
+    if [[ -z "$JOB_ID" ]]; then
+        echo "Slurm allocation succeeded but no job ID was found for $RUNNER_NAME" >&2
+        exit 1
+    fi
+
     srun --jobid=$JOB_ID \
         --mpi=none \
         --container-image=$SQUASH_FILE \