diff --git a/.github/configs/amd-master.yaml b/.github/configs/amd-master.yaml
index 1b1d2e85b..2528798b1 100644
--- a/.github/configs/amd-master.yaml
+++ b/.github/configs/amd-master.yaml
@@ -2846,3 +2846,28 @@ minimaxm3-fp8-mi355x-vllm-mtp:
       - { tp: 8, ep: 8, conc-start: 1, conc-end: 256, spec-decoding: mtp }
       - { tp: 4, conc-start: 1, conc-end: 64, spec-decoding: mtp }
       - { tp: 8, ep: 8, dp-attn: true, conc-start: 128, conc-end: 256, spec-decoding: mtp }
+
+# MiniMax-M3 MXFP8 MI300X day-zero recipe. Reuse the dedicated ROCm image and
+# MI355X serving shape, but retain the default BF16 KV cache because this
+# checkpoint lacks calibrated ROCm FP8 attention scales. Use the TP8-only H100
+# search space: TP8 for latency and TP8+EP8 (TEP) at high concurrency.
+minimaxm3-fp8-mi300x-vllm:
+  image: vllm/vllm-openai-rocm:minimax-m3
+  model: MiniMaxAI/MiniMax-M3-MXFP8
+  model-prefix: minimaxm3
+  runner: mi300x
+  precision: fp8
+  framework: vllm
+  multinode: false
+  scenarios:
+    fixed-seq-len:
+    - isl: 1024
+      osl: 1024
+      search-space:
+      - { tp: 8, conc-start: 4, conc-end: 128 }
+      - { tp: 8, ep: 8, conc-start: 256, conc-end: 256 }
+    - isl: 8192
+      osl: 1024
+      search-space:
+      - { tp: 8, conc-start: 4, conc-end: 64 }
+      - { tp: 8, ep: 8, conc-start: 128, conc-end: 256 }
diff --git a/benchmarks/single_node/fixed_seq_len/minimaxm3_fp8_mi300x.sh b/benchmarks/single_node/fixed_seq_len/minimaxm3_fp8_mi300x.sh
new file mode 100755
index 000000000..e3522e00a
--- /dev/null
+++ b/benchmarks/single_node/fixed_seq_len/minimaxm3_fp8_mi300x.sh
@@ -0,0 +1,88 @@
+#!/usr/bin/env bash
+
+# MiniMax-M3 MXFP8 MI300X (gfx942) single-node vLLM recipe.
+# Reuses the dedicated ROCm image and the MI355X serving shape. Block size 128
+# is mandatory for MSA sparse attention. Keep the default BF16 KV cache on
+# gfx942: the checkpoint has no calibrated q/prob scales for ROCm FP8
+# attention, and vLLM's fallback scale of 1.0 corrupts model accuracy.
+
+source "$(dirname "$0")/../../benchmark_lib.sh"
+
+check_env_vars \
+    MODEL \
+    TP \
+    EP_SIZE \
+    DP_ATTENTION \
+    CONC \
+    ISL \
+    OSL \
+    MAX_MODEL_LEN \
+    RANDOM_RANGE_RATIO \
+    RESULT_FILENAME
+
+if [[ -n "$SLURM_JOB_ID" ]]; then
+  echo "JOB $SLURM_JOB_ID running on $SLURMD_NODENAME"
+fi
+
+if [[ "$MODEL" != /* ]]; then hf download "$MODEL"; fi
+
+if [ -n "$ROCR_VISIBLE_DEVICES" ]; then
+    export HIP_VISIBLE_DEVICES="$ROCR_VISIBLE_DEVICES"
+fi
+
+SERVER_LOG=/workspace/server.log
+export VLLM_ENGINE_READY_TIMEOUT_S=3600
+
+if [ "${EVAL_ONLY}" = "true" ]; then
+    setup_eval_context
+fi
+
+PARALLEL_ARGS=(--tensor-parallel-size "$TP")
+if [ "${DP_ATTENTION}" = "true" ]; then
+    PARALLEL_ARGS=(
+        --tensor-parallel-size 1
+        --data-parallel-size "$TP"
+        --enable-expert-parallel
+    )
+elif [ "$EP_SIZE" -gt 1 ]; then
+    PARALLEL_ARGS+=(--enable-expert-parallel)
+fi
+
+start_gpu_monitor
+
+set -x
+vllm serve "$MODEL" --port "$PORT" \
+    "${PARALLEL_ARGS[@]}" \
+    --block-size 128 \
+    --no-enable-prefix-caching \
+    --language-model-only \
+    --max-model-len "$MAX_MODEL_LEN" \
+    --attention-backend TRITON_ATTN \
+    --enforce-eager \
+    --tool-call-parser minimax_m3 \
+    --reasoning-parser minimax_m3 \
+    --enable-auto-tool-choice > "$SERVER_LOG" 2>&1 &
+
+SERVER_PID=$!
+wait_for_server_ready --port "$PORT" --server-log "$SERVER_LOG" --server-pid "$SERVER_PID"
+
+run_benchmark_serving \
+    --model "$MODEL" \
+    --port "$PORT" \
+    --backend vllm \
+    --input-len "$ISL" \
+    --output-len "$OSL" \
+    --random-range-ratio "$RANDOM_RANGE_RATIO" \
+    --num-prompts "$((CONC * 10))" \
+    --max-concurrency "$CONC" \
+    --result-filename "$RESULT_FILENAME" \
+    --result-dir /workspace/ \
+    --trust-remote-code
+
+if [ "${RUN_EVAL}" = "true" ]; then
+    run_eval --framework lm-eval --port "$PORT"
+    append_lm_eval_summary
+fi
+
+stop_gpu_monitor
+set +x
diff --git a/perf-changelog.yaml b/perf-changelog.yaml
index 57cd2d339..6b16a41c5 100644
--- a/perf-changelog.yaml
+++ b/perf-changelog.yaml
@@ -3735,3 +3735,13 @@
     - "Layouts: TP8 / TP4 (latency), TP8+EP8 / TP4+EP4 (TEP), TP8+EP8 dp-attn (DEP) across 1k1k and 8k1k — non-MTP search space trimmed at the extreme-concurrency end, tp2-ep2 dropped, mirroring the minimaxm3-fp8-b300-vllm-mtp search space"
     - "[AI generated draft test] The shipped ROCm image's AMD MiniMax-M3 model lacks SupportsEagle3, so the recipe patches it in-place at runtime (functionstackx/vllm#1, ported from nvidia/model.py) before serving — validates EAGLE3 on MI355X ahead of an image rebuild"
   pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/1745
+
+- config-keys:
+    - minimaxm3-fp8-mi300x-vllm
+  description:
+    - "Initial submission: MiniMax-M3 MXFP8 day-zero single-node vLLM benchmark on MI300X / gfx942"
+    - "Image and serving shape reuse the MI355X recipe: vllm/vllm-openai-rocm:minimax-m3, block size 128, prefix caching disabled, TRITON_ATTN, language-model-only, and eager execution"
+    - "Use the default BF16 KV cache on MI300X because MiniMax-M3-MXFP8 lacks calibrated q/prob scales for ROCm FP8 attention; vLLM warns that its fallback scale of 1.0 may cause accuracy issues"
+    - "H100-aligned layouts and concurrency ranges: TP8 and TP8+EP8 across 1k1k and 8k1k"
+    - "Fix launch_mi300x-amds.sh node exclusion to use the current short Slurm node name"
+  pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/1746
diff --git a/runners/launch_mi300x-amds.sh b/runners/launch_mi300x-amds.sh
index 47ad12698..ce04ceadd 100644
--- a/runners/launch_mi300x-amds.sh
+++ b/runners/launch_mi300x-amds.sh
@@ -10,8 +10,8 @@ LOCK_FILE="${SQUASH_FILE}.lock"
 set -x
 
 # Exclude known-bad nodes; let Slurm pick from anything else:
-#   chi-mi300x-049: drained (persistent /nvme_home disk-full)
-JOB_ID=$(salloc --partition=$PARTITION --exclude=chi-mi300x-049.ord.vultr.cpe.ice.amd.com --gres=gpu:$TP --cpus-per-task=256 --time=180 --no-shell --job-name="$RUNNER_NAME" 2>&1 | tee /dev/stderr | grep -oP 'Granted job allocation \K[0-9]+')
+#   chi-mi300x-049: persistent /nvme_home disk-full
+JOB_ID=$(salloc --partition=$PARTITION --exclude=chi-mi300x-049 --gres=gpu:$TP --cpus-per-task=256 --time=180 --no-shell --job-name="$RUNNER_NAME" 2>&1 | tee /dev/stderr | grep -oP 'Granted job allocation \K[0-9]+')
 
 if [ -z "$JOB_ID" ]; then
     echo "ERROR: salloc failed to allocate a job"
@@ -31,7 +31,7 @@ srun --jobid=$JOB_ID --job-name="$RUNNER_NAME" bash -c "
 "
 srun --jobid=$JOB_ID \
 --container-image=$SQUASH_FILE \
---container-mounts=$GITHUB_WORKSPACE:/workspace/,$HF_HUB_CACHE_MOUNT:$HF_HUB_CACHE \
+--container-mounts=$GITHUB_WORKSPACE:/workspace/,$HF_HUB_CACHE_MOUNT:$HF_HUB_CACHE,/dev/kfd:/dev/kfd,/dev/dri:/dev/dri \
 --container-mount-home \
 --container-writable \
 --container-remap-root \
@@ -39,4 +39,4 @@ srun --jobid=$JOB_ID \
 --no-container-entrypoint --export=ALL \
 bash benchmarks/single_node/${SCENARIO_SUBDIR}${EXP_NAME%%_*}_${PRECISION}_mi300x.sh
 
-scancel $JOB_ID
\ No newline at end of file
+scancel $JOB_ID