diff --git a/benchmarks/kernels/benchmark_host_buffer_kv_copy.py b/benchmarks/kernels/benchmark_host_buffer_kv_copy.py
new file mode 100644
index 000000000000..be4051c0bd41
--- /dev/null
+++ b/benchmarks/kernels/benchmark_host_buffer_kv_copy.py
@@ -0,0 +1,256 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+"""Benchmark the NIXL host-buffer KV copies across KV cache groups.
+
+`NixlBaseConnectorWorker.sync_recved_kv_to_device` / `save_kv_to_host` move a
+request's blocks between the CPU transfer buffer and the device KV cache. Block
+ids are unique across KV cache groups, so the per-group copies can be issued as
+one. This measures what that coalescing saves: the copied bytes are identical,
+only the number of launches changes.
+"""
+
+import time
+
+import torch
+from tabulate import tabulate
+
+from vllm.distributed.kv_transfer.kv_connector.utils import copy_kv_blocks
+from vllm.logger import init_logger
+from vllm.utils.argparse_utils import FlexibleArgumentParser
+from vllm.utils.torch_utils import STR_DTYPE_TO_TORCH_DTYPE
+
+logger = init_logger(__name__)
+
+
+def _make_caches(
+    num_layers: int,
+    num_blocks: int,
+    block_size: int,
+    num_heads: int,
+    head_size: int,
+    dtype: torch.dtype,
+    device: str,
+) -> tuple[dict[str, torch.Tensor], dict[str, torch.Tensor]]:
+    """Host transfer buffers and device KV caches, shaped as the worker builds them."""
+    shape = (num_blocks, block_size, num_heads, head_size)
+    host = {
+        f"layer.{i}": torch.randn(shape, dtype=dtype, device="cpu")
+        for i in range(num_layers)
+    }
+    device_caches = {
+        f"layer.{i}": torch.zeros(shape, dtype=dtype, device=device)
+        for i in range(num_layers)
+    }
+    return host, device_caches
+
+
+def _split_into_groups(block_ids: list[int], num_groups: int) -> list[list[int]]:
+    """Partition block ids into groups, mirroring the global BlockPool id space."""
+    per_group = len(block_ids) // num_groups
+    return [block_ids[i * per_group : (i + 1) * per_group] for i in range(num_groups)]
+
+
+def _percentile(sorted_values: list[float], q: float) -> float:
+    return sorted_values[min(len(sorted_values) - 1, int(len(sorted_values) * q))]
+
+
+@torch.inference_mode()
+def _time_paired(
+    host: dict[str, torch.Tensor],
+    device_caches: dict[str, torch.Tensor],
+    group_block_ids: list[list[int]],
+    direction: str,
+    num_iters: int,
+) -> tuple[list[float], list[float]]:
+    """Per-group and coalesced latencies, measured alternately.
+
+    Interleaving the two variants keeps drift in machine state from landing on
+    one of them: each iteration times both, back to back.
+    """
+    src, dst = (host, device_caches) if direction == "h2d" else (device_caches, host)
+    merged = [[b for group in group_block_ids for b in group]]
+
+    def _run(batches: list[list[int]]) -> float:
+        start = time.perf_counter()
+        for ids in batches:
+            copy_kv_blocks(src, dst, ids, ids, direction)
+        torch.accelerator.synchronize()
+        return time.perf_counter() - start
+
+    for _ in range(5):  # warmup
+        _run(group_block_ids)
+        _run(merged)
+
+    per_group: list[float] = []
+    coalesced: list[float] = []
+    for i in range(num_iters):
+        # Alternate which variant goes first so ordering cannot favour either.
+        if i % 2:
+            per_group.append(_run(group_block_ids))
+            coalesced.append(_run(merged))
+        else:
+            coalesced.append(_run(merged))
+            per_group.append(_run(group_block_ids))
+    return sorted(per_group), sorted(coalesced)
+
+
+@torch.inference_mode()
+def _count_device_ops(
+    host: dict[str, torch.Tensor],
+    device_caches: dict[str, torch.Tensor],
+    group_block_ids: list[list[int]],
+    direction: str,
+    coalesced: bool,
+) -> dict[str, int]:
+    """Device-op counts for one request-sync, via the profiler.
+
+    Counts do not depend on how busy the machine is, so this is the part of
+    the comparison that reproduces exactly.
+    """
+    from torch.profiler import ProfilerActivity, profile
+
+    src, dst = (host, device_caches) if direction == "h2d" else (device_caches, host)
+    batches = (
+        [[b for group in group_block_ids for b in group]]
+        if coalesced
+        else group_block_ids
+    )
+
+    for ids in batches:  # warm up allocator/autograd caches
+        copy_kv_blocks(src, dst, ids, ids, direction)
+    torch.accelerator.synchronize()
+
+    with profile(activities=[ProfilerActivity.CPU, ProfilerActivity.CUDA]) as prof:
+        for ids in batches:
+            copy_kv_blocks(src, dst, ids, ids, direction)
+        torch.accelerator.synchronize()
+
+    counts: dict[str, int] = {}
+    for evt in prof.key_averages():
+        if evt.count and (evt.self_device_time_total or evt.device_time_total):
+            counts[evt.key] = counts.get(evt.key, 0) + evt.count
+    return counts
+
+
+def _report_op_counts(host, device_caches, block_ids, groups):
+    print("\nDevice-op counts per request-sync (profiler; independent of load):")
+    rows = []
+    for num_groups in groups:
+        group_block_ids = _split_into_groups(block_ids, num_groups)
+        for direction in ("h2d", "d2h"):
+            before = _count_device_ops(
+                host, device_caches, group_block_ids, direction, coalesced=False
+            )
+            after = _count_device_ops(
+                host, device_caches, group_block_ids, direction, coalesced=True
+            )
+            for key in sorted(set(before) | set(after)):
+                b, a = before.get(key, 0), after.get(key, 0)
+                if b != a:
+                    rows.append([num_groups, direction, key, b, a])
+    print(
+        tabulate(
+            rows,
+            headers=["groups", "direction", "op", "per-group", "coalesced"],
+        )
+    )
+
+
+def main(args):
+    dtype = STR_DTYPE_TO_TORCH_DTYPE[args.dtype]
+    torch.manual_seed(args.seed)
+
+    host, device_caches = _make_caches(
+        num_layers=args.num_layers,
+        num_blocks=args.num_blocks,
+        block_size=args.block_size,
+        num_heads=args.num_heads,
+        head_size=args.head_size,
+        dtype=dtype,
+        device="cuda",
+    )
+    block_ids = list(range(args.blocks_per_request))
+
+    rows = []
+    for num_groups in args.groups:
+        group_block_ids = _split_into_groups(block_ids, num_groups)
+        for direction in ("h2d", "d2h"):
+            per_group, coalesced = _time_paired(
+                host,
+                device_caches,
+                group_block_ids,
+                direction,
+                num_iters=args.iters,
+            )
+            # Interference only ever adds time, so the minimum is the most
+            # stable estimator here; the p10 spread above it says how noisy
+            # the machine was while sampling.
+            best_before = per_group[0]
+            best_after = coalesced[0]
+            spread = (_percentile(coalesced, 0.5) - best_after) / best_after
+            rows.append(
+                [
+                    num_groups,
+                    direction,
+                    best_before * 1e6,
+                    best_after * 1e6,
+                    (best_before - best_after) / best_before * 100.0,
+                    spread * 100.0,
+                ]
+            )
+
+    print(
+        f"layers={args.num_layers} num_blocks={args.num_blocks} "
+        f"block_size={args.block_size} heads={args.num_heads} "
+        f"head_size={args.head_size} dtype={args.dtype} "
+        f"blocks/request={args.blocks_per_request} iters={args.iters}"
+    )
+    print(
+        tabulate(
+            rows,
+            headers=[
+                "groups",
+                "direction",
+                "per-group min (µs)",
+                "coalesced min (µs)",
+                "saved (%)",
+                "p50 over min (%)",
+            ],
+            floatfmt=".3f",
+        )
+    )
+
+    if args.op_counts:
+        _report_op_counts(host, device_caches, block_ids, args.groups)
+
+
+if __name__ == "__main__":
+    parser = FlexibleArgumentParser()
+    parser.add_argument("--num-layers", type=int, default=48)
+    parser.add_argument("--num-blocks", type=int, default=2048)
+    parser.add_argument("--block-size", type=int, choices=[16, 32], default=16)
+    parser.add_argument("--num-heads", type=int, default=8)
+    parser.add_argument("--head-size", type=int, default=128)
+    parser.add_argument("--blocks-per-request", type=int, default=60)
+    parser.add_argument(
+        "--groups",
+        type=int,
+        nargs="+",
+        default=[1, 2, 3, 6],
+        help="KV cache group counts to sweep (1 = non-hybrid model).",
+    )
+    parser.add_argument(
+        "--dtype",
+        type=str,
+        choices=["half", "bfloat16", "float"],
+        default="bfloat16",
+    )
+    parser.add_argument("--iters", type=int, default=100)
+    parser.add_argument(
+        "--op-counts",
+        action="store_true",
+        help="Also report profiler op counts, which do not vary with load.",
+    )
+    parser.add_argument("--seed", type=int, default=0)
+
+    main(parser.parse_args())
diff --git a/tests/v1/kv_connector/unit/test_nixl_connector_hma.py b/tests/v1/kv_connector/unit/test_nixl_connector_hma.py
index 629e8c7ecc3f..cb6c601ba311 100644
--- a/tests/v1/kv_connector/unit/test_nixl_connector_hma.py
+++ b/tests/v1/kv_connector/unit/test_nixl_connector_hma.py
@@ -2396,3 +2396,50 @@ def test_push_write_hybrid_mla_replicates_attention():
         assert spec.remote_block_ids == [[7, 8], [3]]
         assert call.kwargs["local_xfer_side_handle"] == local_handle
         assert call.kwargs["remote_xfer_side_handle"] == remote_handle
+
+
+def _make_host_buffer_worker(copy_op):
+    """A worker stripped down to what the host-buffer copy paths touch."""
+    from vllm.distributed.kv_transfer.kv_connector.v1.nixl.worker import (
+        NixlConnectorWorker,
+    )
+
+    worker = object.__new__(NixlConnectorWorker)
+    worker.use_host_buffer = True
+    worker.copy_blocks = copy_op
+    return worker
+
+
+@pytest.mark.cpu_test
+@pytest.mark.parametrize(
+    "group_block_ids,expected_ids",
+    [
+        # Single group (non-hybrid model): one copy, as before.
+        ([[1, 2, 3]], [1, 2, 3]),
+        # Hybrid model: three groups collapse into a single copy.
+        ([[1, 2], [3, 4], [5]], [1, 2, 3, 4, 5]),
+        # An empty group contributes no ids.
+        ([[1, 2], []], [1, 2]),
+    ],
+)
+def test_sync_recved_kv_issues_one_copy_per_request(group_block_ids, expected_ids):
+    """h2d copies are issued once per request, not once per KV cache group."""
+    from vllm.distributed.kv_transfer.kv_connector.v1.nixl.metadata import ReqMeta
+
+    calls = []
+    worker = _make_host_buffer_worker(
+        lambda src, dst, src_ids, dst_ids, direction: calls.append(
+            (src_ids, dst_ids, direction)
+        )
+    )
+    worker.host_xfer_buffers = {"layer": None}
+    worker.device_kv_caches = {"layer": None}
+
+    meta = ReqMeta(
+        local_block_ids=group_block_ids,
+        local_physical_block_ids=group_block_ids,
+        tp_size=1,
+    )
+    worker.sync_recved_kv_to_device("req", meta)
+
+    assert calls == [(expected_ids, expected_ids, "h2d")]
diff --git a/vllm/distributed/kv_transfer/kv_connector/v1/nixl/base_worker.py b/vllm/distributed/kv_transfer/kv_connector/v1/nixl/base_worker.py
index 0cbdd786d8c3..48bb5edc7fcc 100644
--- a/vllm/distributed/kv_transfer/kv_connector/v1/nixl/base_worker.py
+++ b/vllm/distributed/kv_transfer/kv_connector/v1/nixl/base_worker.py
@@ -2702,17 +2702,18 @@ def sync_recved_kv_to_device(self, req_id: str, meta: ReqMeta):
         assert self.copy_blocks is not None
 
         local_block_ids = meta.local_physical_block_ids
-        # TODO (NickLucche) D2H<>H2D ops could benefit from coalescing io across groups
-        # The h2d block copies below are intentionally synchronous.
+        # Every KV cache group allocates from the same BlockPool, so block ids
+        # are unique across groups and the per-group copies can be issued as one.
+        block_ids = [block_id for group in local_block_ids for block_id in group]
+        # The h2d block copy below is intentionally synchronous.
         with gpu_sync_allowed():
-            for group_block_ids in local_block_ids:
-                self.copy_blocks(
-                    self.host_xfer_buffers,
-                    self.device_kv_caches,
-                    group_block_ids,
-                    group_block_ids,
-                    "h2d",
-                )
+            self.copy_blocks(
+                self.host_xfer_buffers,
+                self.device_kv_caches,
+                block_ids,
+                block_ids,
+                "h2d",
+            )
         if logger.isEnabledFor(logging.DEBUG):
             logger.debug(
                 "synced recved kv of request[%s] to device kv buffer,"
@@ -2740,14 +2741,18 @@ def save_kv_to_host(self, metadata: NixlConnectorMetadata):
                         ",".join(map(str, meta.local_physical_block_ids)),
                     )
                 # blocking
-                for group_block_ids in meta.local_physical_block_ids:
-                    self.copy_blocks(
-                        self.device_kv_caches,
-                        self.host_xfer_buffers,
-                        group_block_ids,
-                        group_block_ids,
-                        "d2h",
-                    )
+                block_ids = [
+                    block_id
+                    for group in meta.local_physical_block_ids
+                    for block_id in group
+                ]
+                self.copy_blocks(
+                    self.device_kv_caches,
+                    self.host_xfer_buffers,
+                    block_ids,
+                    block_ids,
+                    "d2h",
+                )
 
     @cached_property
     def _attention_kv_caches(self) -> list[torch.Tensor]: