diff --git a/benchmarks/single_node/agentic/qwen3.5_fp4_b300_sglang_mtp.sh b/benchmarks/single_node/agentic/qwen3.5_fp4_b300_sglang_mtp.sh
new file mode 100755
index 0000000000..03aa654219
--- /dev/null
+++ b/benchmarks/single_node/agentic/qwen3.5_fp4_b300_sglang_mtp.sh
@@ -0,0 +1,177 @@
+#!/usr/bin/env bash
+set -euo pipefail
+set -x
+
+# AgentX trace replay for Qwen3.5-397B-A17B NVFP4 on B300 with SGLang
+# native NEXTN MTP. Throughput uses the committed golden synthetic AL; evals
+# retain real target-model verification.
+
+source "$(dirname "$0")/../../benchmark_lib.sh"
+
+check_env_vars \
+    MODEL TP CONC EP_SIZE KV_OFFLOADING \
+    TOTAL_CPU_DRAM_GB RESULT_DIR DURATION
+
+SCHEDULER_RECV_INTERVAL=${SCHEDULER_RECV_INTERVAL:-10}
+
+if [[ -n "${SLURM_JOB_ID:-}" ]]; then
+    echo "JOB $SLURM_JOB_ID running on ${SLURMD_NODENAME:-unknown}"
+fi
+
+if [[ -n "${MODEL_PATH:-}" ]]; then
+    if [[ ! -d "$MODEL_PATH" || -z "$(ls -A "$MODEL_PATH" 2>/dev/null)" ]]; then
+        hf download "$MODEL" --local-dir "$MODEL_PATH"
+    fi
+else
+    hf download "$MODEL"
+    export MODEL_PATH="$MODEL"
+fi
+nvidia-smi
+
+export WEKA_LOADER_OVERRIDE=semianalysis_cc_traces_weka_062126_256k
+resolve_trace_source
+install_agentic_deps
+
+SERVER_LOG="$RESULT_DIR/server.log"
+mkdir -p "$RESULT_DIR"
+
+CACHE_ARGS=()
+if require_agentic_kv_offload_backend hicache; then
+    REQUESTED_HICACHE_TOTAL_GB="${HICACHE_TOTAL_CPU_DRAM_GB:-$TOTAL_CPU_DRAM_GB}"
+    if [ "$REQUESTED_HICACHE_TOTAL_GB" -gt "$TOTAL_CPU_DRAM_GB" ]; then
+        echo "Error: requested HiCache pool ${REQUESTED_HICACHE_TOTAL_GB} GB exceeds configured capacity ${TOTAL_CPU_DRAM_GB} GB" >&2
+        exit 1
+    fi
+    TOTAL_CPU_DRAM_GB="$REQUESTED_HICACHE_TOTAL_GB"
+    # SGLang applies --hicache-size independently to Qwen's target KV and
+    # Mamba pools. Native NEXTN also creates a draft KV pool with the same
+    # slot count; its one attention layer adds 1/15 of the target KV bytes.
+    # Reserve 1 GB/rank for page alignment and enforce H * 31/15 per rank.
+    HICACHE_ALIGNMENT_RESERVE_GB=$TP
+    HICACHE_USABLE_TOTAL_GB=$((TOTAL_CPU_DRAM_GB - HICACHE_ALIGNMENT_RESERVE_GB))
+    if [ "$HICACHE_USABLE_TOTAL_GB" -lt 1 ]; then
+        echo "Error: insufficient DRAM after HiCache alignment reserve" >&2
+        exit 1
+    fi
+    MAX_HICACHE_SIZE_GB=$((HICACHE_USABLE_TOTAL_GB * 15 / TP / 31))
+    HICACHE_SIZE_GB="${HICACHE_SIZE_GB:-$MAX_HICACHE_SIZE_GB}"
+    if [ "$HICACHE_SIZE_GB" -lt 1 ] || [ "$HICACHE_SIZE_GB" -gt "$MAX_HICACHE_SIZE_GB" ]; then
+        echo "Error: HICACHE_SIZE_GB=$HICACHE_SIZE_GB outside 1..$MAX_HICACHE_SIZE_GB" >&2
+        exit 1
+    fi
+    PROJECTED_HICACHE_TOTAL_GB=$(((HICACHE_SIZE_GB * TP * 31 + 14) / 15 + HICACHE_ALIGNMENT_RESERVE_GB))
+    if [ "$PROJECTED_HICACHE_TOTAL_GB" -gt "$TOTAL_CPU_DRAM_GB" ]; then
+        echo "Error: projected HiCache use ${PROJECTED_HICACHE_TOTAL_GB} GB exceeds configured capacity ${TOTAL_CPU_DRAM_GB} GB" >&2
+        exit 1
+    fi
+    echo "HiCache CPU pools: ${HICACHE_SIZE_GB} GB target + Mamba + 1/15 draft per rank across TP=${TP}; projected node total ${PROJECTED_HICACHE_TOTAL_GB} GB <= ${TOTAL_CPU_DRAM_GB} GB"
+    CACHE_ARGS=(
+        --page-size 64
+        --enable-hierarchical-cache
+        --hicache-size "$HICACHE_SIZE_GB"
+        --hicache-io-backend kernel
+        --hicache-mem-layout page_first
+        --hicache-write-policy write_through_selective
+    )
+fi
+
+PARALLEL_ARGS=(
+    --tp "$TP"
+    --dp 1
+    --ep-size "$EP_SIZE"
+)
+STREAM_INTERVAL=50
+
+# TP4 needs parallel tokenization to keep 256k AgentX warmups below the client
+# request timeout. Keep TP2 on SGLang's single-worker default: multi-tokenizer
+# startup races with the TP2 HiCache shared-memory initialization path.
+TOKENIZER_ARGS=()
+if [ "$TP" -ge 4 ]; then
+    TOKENIZER_ARGS=(--tokenizer-worker-num 6)
+fi
+
+# AgentX concurrency counts live session trees rather than individual HTTP
+# requests. Leave room for subagent fan-out and avoid spending HBM on graphs
+# above the batch sizes that remain useful for this long-context workload.
+MAX_RUNNING_REQUESTS=$((2 * CONC))
+CUDA_GRAPH_MAX_BS="$CONC"
+[ "$CUDA_GRAPH_MAX_BS" -gt 64 ] && CUDA_GRAPH_MAX_BS=64
+
+export TORCH_CUDA_ARCH_LIST="10.0"
+export PYTHONNOUSERSITE=1
+export NCCL_NVLS_ENABLE=1
+export SGL_ENABLE_JIT_DEEPGEMM=false
+export SGLANG_ENABLE_FLASHINFER_GEMM=true
+# Keep server-side connections alive beyond AIPerf's 300-second client pool
+# timeout so bursty AgentX trajectories cannot reuse a closing idle socket.
+export SGLANG_TIMEOUT_KEEP_ALIVE=1800
+
+if [ "${EVAL_ONLY:-false}" != "true" ]; then
+    export SGLANG_SIMULATE_ACC_LEN=3.39
+    export SGLANG_SIMULATE_ACC_METHOD=match-expected
+    export SGLANG_SIMULATE_ACC_TOKEN_MODE=real-draft-token
+fi
+
+SGLANG_CMD=(
+    python3 -m sglang.launch_server
+    --model-path "$MODEL_PATH"
+    --served-model-name "$MODEL"
+    --host 0.0.0.0
+    --port "$PORT"
+    --trust-remote-code
+    "${PARALLEL_ARGS[@]}"
+    --enable-symm-mem
+    --quantization modelopt_fp4
+    --fp4-gemm-backend flashinfer_cutlass
+    --kv-cache-dtype fp8_e4m3
+    --mamba-ssm-dtype bfloat16
+    --attention-backend trtllm_mha
+    --moe-runner-backend flashinfer_trtllm
+    --cuda-graph-max-bs "$CUDA_GRAPH_MAX_BS"
+    --max-running-requests "$MAX_RUNNING_REQUESTS"
+    --max-prefill-tokens 16384
+    --chunked-prefill-size 16384
+    --mem-fraction-static 0.80
+    --stream-interval "$STREAM_INTERVAL"
+    --scheduler-recv-interval "$SCHEDULER_RECV_INTERVAL"
+    "${TOKENIZER_ARGS[@]}"
+    --tokenizer-path "$MODEL"
+    --reasoning-parser qwen3
+    --tool-call-parser qwen3_coder
+    --speculative-algorithm NEXTN
+    --speculative-num-steps 3
+    --speculative-eagle-topk 1
+    --speculative-num-draft-tokens 4
+    --enable-metrics
+    --enable-cache-report
+    "${CACHE_ARGS[@]}"
+)
+
+printf '%q ' "${SGLANG_CMD[@]}" | tee "$RESULT_DIR/sglang_command.txt"
+printf '\n' | tee -a "$RESULT_DIR/sglang_command.txt"
+"${SGLANG_CMD[@]}" > "$SERVER_LOG" 2>&1 &
+SERVER_PID=$!
+
+capture_cache_metrics() {
+    {
+        echo "=== SGLang cache metrics snapshot $(date --iso-8601=seconds) ==="
+        curl -fsS "http://localhost:$PORT/metrics" 2>/dev/null \
+            | grep -E '^(sglang:(cache_hit_rate|cached_tokens_total|prompt_tokens_total|hicache_host_used_tokens|hicache_host_total_tokens|token_usage|num_requests_running|num_requests_waiting))' \
+            || true
+        echo "============================================================"
+    } >> "$SERVER_LOG"
+}
+
+wait_for_server_ready --port "$PORT" --server-log "$SERVER_LOG" --server-pid "$SERVER_PID"
+
+capture_cache_metrics
+trap capture_cache_metrics EXIT
+
+if [ "${EVAL_ONLY:-false}" = "true" ]; then
+    run_eval --port "$PORT"
+else
+    build_replay_cmd "$RESULT_DIR"
+    REPLAY_CMD+=" --trace-idle-gap-cap-seconds 300"
+    REPLAY_CMD+=" --server-metrics http://localhost:$PORT/metrics"
+    run_agentic_replay_and_write_outputs "$RESULT_DIR"
+fi
diff --git a/configs/nvidia-master.yaml b/configs/nvidia-master.yaml
index a30af59641..8c2c11c050 100644
--- a/configs/nvidia-master.yaml
+++ b/configs/nvidia-master.yaml
@@ -7284,6 +7284,35 @@ qwen3.5-fp8-b300-sglang-agentic-hicache:
       - { tp: 4, ep: 1, kv-offloading: none, conc-list: [1, 2, 4, 8, 16, 32] }
       - { tp: 4, ep: 1, kv-offloading: dram, kv-offload-backend: { name: hicache }, conc-list: [16, 32, 48, 64] }
 
+# Qwen3.5 target FP8 KV is 15,360 bytes/token before the native MTP layer.
+# TP4 replicates the model's two KV heads and measured 30.5 KB/token node-wide;
+# TP2 measured 15.3 KB/token. Against the 256k trace's 88.8k median and 148.8k
+# p75 ISL, TP2 no-offload peaks at c32. HiCache overlaps from c32; probes put
+# the throughput transition between c34 and c40, while host KV reaches 98% at
+# c52 and c56 is the last useful stress point. c64+ warmups only deepen the
+# queue and are excluded. TP4 no-offload remains healthy into the c56-72
+# neighborhood, which is sampled every four requests around the cache cliff.
+# TP4 HiCache is dominated there: c68 sustained 197k tok/s versus 224k for
+# no-offload c64, while c76 fell to 158k with host KV full. Clean HiCache c94
+# was also dominated by no-offload c56 (210k vs 234k tok/s, with 44 ms vs 7 ms
+# median ITL), so the TP4 HiCache arm is excluded.
+# DEP8 was rejected after fragmenting the trace across eight radix caches.
+qwen3.5-fp4-b300-sglang-agentic-mtp:
+  image: lmsysorg/sglang:v0.5.16-cu130
+  model: nvidia/Qwen3.5-397B-A17B-NVFP4
+  model-prefix: qwen3.5
+  runner: cluster:b300-nv
+  precision: fp4
+  framework: sglang
+  multinode: false
+  scenarios:
+    agentic-coding:
+    - dram-utilization: 0.80
+      search-space:
+      - { tp: 4, ep: 1, spec-decoding: mtp, kv-offloading: none, conc-list: [1, 4, 8, 12, 16, 20, 24, 28, 32, 40, 48, 56, 60, 64, 68, 72] }
+      - { tp: 2, ep: 2, spec-decoding: mtp, kv-offloading: none, conc-list: [1, 4, 8, 12, 16, 20, 24, 28, 32] }
+      - { tp: 2, ep: 2, spec-decoding: mtp, kv-offloading: dram, kv-offload-backend: { name: hicache }, conc-list: [32, 34, 36, 38, 40, 44, 48, 52, 56] }
+
 kimik2.5-fp4-b200-vllm-agentic-lmcache:
   image: vllm/vllm-openai:v0.22.0
   model: nvidia/Kimi-K2.5-NVFP4
diff --git a/perf-changelog.yaml b/perf-changelog.yaml
index 6eebd97070..a62f84e01b 100644
--- a/perf-changelog.yaml
+++ b/perf-changelog.yaml
@@ -5355,3 +5355,12 @@
     - "Apply the accuracy-gated Kimi-K2.5 MXFP4 settings: tuned AITER MXFP4 MoE, fused shared experts, FP8 KV cache, block size 16, 16384 batched tokens, 512 sequences, async scheduling, gpu-memory-utilization 0.85 (headroom for CUDA-graph capture on MI355X), and the AITER BF16 GEMM path"
     - "Extend the TP4 and TP8 8k1k concurrency sweep from 64 to 128 (1k1k deprecated per #2263)"
   pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/2213
+
+- config-keys:
+    - qwen3.5-fp4-b300-sglang-agentic-mtp
+  description:
+    - "Add Qwen3.5-397B-A17B NVFP4 AgentX benchmark on B300 with SGLang native NEXTN MTP"
+    - "Use the 256k trace dataset and golden synthetic acceptance length 3.39"
+    - "Pin AIPerf trace idle-gap support and cap per-trace idle gaps at 300 seconds"
+    - "Image: lmsysorg/sglang:v0.5.16-cu130"
+  pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/2421
diff --git a/utils/aiperf b/utils/aiperf
index be758d6218..c12846297f 160000
--- a/utils/aiperf
+++ b/utils/aiperf
@@ -1 +1 @@
-Subproject commit be758d6218268171e2957fbec9d4f557275bca2d
+Subproject commit c12846297fb5b9057cbbd6c8cf55637d8353385d