diff --git a/benchmarks/multi_node/srt-slurm-recipes/configs/minimax-m3-gb200-vllm-fixes.sh b/benchmarks/multi_node/srt-slurm-recipes/configs/minimax-m3-gb200-vllm-fixes.sh
deleted file mode 100755
index c0eed0a517..0000000000
--- a/benchmarks/multi_node/srt-slurm-recipes/configs/minimax-m3-gb200-vllm-fixes.sh
+++ /dev/null
@@ -1,38 +0,0 @@
-#!/usr/bin/env bash
-set -euo pipefail
-
-python3 - <<'PYEOF'
-from importlib.util import find_spec
-from pathlib import Path
-
-spec = find_spec("vllm")
-if not spec or not spec.origin:
-    raise RuntimeError("vllm is not installed")
-root = Path(spec.origin).parent
-patches = {
-    root / "distributed/device_communicators/flashinfer_all_reduce.py": [
-        (
-            "            comm_backend=comm_backend,\n"
-            "            group=group,\n",
-            "            comm_backend=comm_backend,\n"
-            '            force_oneshot_support=backend == "mnnvl",\n'
-            "            group=group,\n",
-        ),
-    ],
-    root / "models/minimax_m3/nvidia/sparse_attention_msa.py": [
-        (
-            "            prefill_topk = topk[:, nd:num_tokens, :]\n",
-            "            prefill_topk = topk[:, nd:num_tokens, :].contiguous()\n",
-        ),
-    ],
-}
-for path, edits in patches.items():
-    source = path.read_text()
-    for old, new in edits:
-        if new in source:
-            continue
-        if source.count(old) != 1:
-            raise RuntimeError(f"missing or ambiguous patch anchor in {path}")
-        source = source.replace(old, new, 1)
-    path.write_text(source)
-PYEOF
diff --git a/benchmarks/multi_node/srt-slurm-recipes/configs/minimax-m3-gb300-vllm-fixes.sh b/benchmarks/multi_node/srt-slurm-recipes/configs/minimax-m3-gb300-vllm-fixes.sh
deleted file mode 100755
index c0eed0a517..0000000000
--- a/benchmarks/multi_node/srt-slurm-recipes/configs/minimax-m3-gb300-vllm-fixes.sh
+++ /dev/null
@@ -1,38 +0,0 @@
-#!/usr/bin/env bash
-set -euo pipefail
-
-python3 - <<'PYEOF'
-from importlib.util import find_spec
-from pathlib import Path
-
-spec = find_spec("vllm")
-if not spec or not spec.origin:
-    raise RuntimeError("vllm is not installed")
-root = Path(spec.origin).parent
-patches = {
-    root / "distributed/device_communicators/flashinfer_all_reduce.py": [
-        (
-            "            comm_backend=comm_backend,\n"
-            "            group=group,\n",
-            "            comm_backend=comm_backend,\n"
-            '            force_oneshot_support=backend == "mnnvl",\n'
-            "            group=group,\n",
-        ),
-    ],
-    root / "models/minimax_m3/nvidia/sparse_attention_msa.py": [
-        (
-            "            prefill_topk = topk[:, nd:num_tokens, :]\n",
-            "            prefill_topk = topk[:, nd:num_tokens, :].contiguous()\n",
-        ),
-    ],
-}
-for path, edits in patches.items():
-    source = path.read_text()
-    for old, new in edits:
-        if new in source:
-            continue
-        if source.count(old) != 1:
-            raise RuntimeError(f"missing or ambiguous patch anchor in {path}")
-        source = source.replace(old, new, 1)
-    path.write_text(source)
-PYEOF
diff --git a/benchmarks/multi_node/srt-slurm-recipes/configs/minimax-m3-vllm-fixes.sh b/benchmarks/multi_node/srt-slurm-recipes/configs/minimax-m3-vllm-fixes.sh
deleted file mode 100755
index 02862bba39..0000000000
--- a/benchmarks/multi_node/srt-slurm-recipes/configs/minimax-m3-vllm-fixes.sh
+++ /dev/null
@@ -1,47 +0,0 @@
-#!/usr/bin/env bash
-set -euo pipefail
-
-python3 - <<'PYEOF'
-from importlib.util import find_spec
-from pathlib import Path
-
-spec = find_spec("vllm")
-if not spec or not spec.origin:
-    raise RuntimeError("vllm is not installed")
-root = Path(spec.origin).parent
-patches = {
-    root / "models/minimax_m3/nvidia/sparse_attention_msa.py": [
-        (
-            "            prefill_topk = topk[:, nd:num_tokens, :]\n",
-            "            prefill_topk = topk[:, nd:num_tokens, :].contiguous()\n",
-        ),
-    ],
-    root / "distributed/kv_transfer/kv_connector/v1/nixl/base_worker.py": [
-        (
-            "            for i, local_len in enumerate(self.block_len_per_layer):\n",
-            "            total_kv_heads = self.transfer_topo.total_num_kv_heads\n"
-            "            local_heads = self.transfer_topo.local_physical_heads\n"
-            "            remote_heads = max(1, total_kv_heads // remote_tp_size)\n"
-            "            for i, local_len in enumerate(self.block_len_per_layer):\n",
-        ),
-        (
-            "remote_len == (local_len * tp_ratio) // block_size_ratio,",
-            "remote_len == (local_len * remote_heads // local_heads) "
-            "// block_size_ratio,",
-        ),
-        (
-            "remote_len == local_len // (-tp_ratio),",
-            "remote_len == local_len * remote_heads // local_heads,",
-        ),
-    ],
-}
-for path, edits in patches.items():
-    source = path.read_text()
-    for old, new in edits:
-        if new in source:
-            continue
-        if source.count(old) != 1:
-            raise RuntimeError(f"missing or ambiguous patch anchor in {path}")
-        source = source.replace(old, new, 1)
-    path.write_text(source)
-PYEOF
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb200-fp8/8k1k/1p2d-dep4-dep8-8k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb200-fp8/8k1k/1p2d-dep4-dep8-8k1k.yaml
index 4129408ebd..7dd82b6098 100644
--- a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb200-fp8/8k1k/1p2d-dep4-dep8-8k1k.yaml
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb200-fp8/8k1k/1p2d-dep4-dep8-8k1k.yaml
@@ -6,13 +6,13 @@ name: "minimax-m3-vllm-disagg-gb200-1p2d-dep4-dep8-fp8-8k1k"
 
 model:
   path: "minimax-m3-mxfp8"
-  container: "vllm/vllm-openai:minimax-m3-perf-arm64-13.0.1-7a67223"
+  container: "vllm/vllm-openai:nightly-4080263bb2c5d10deac17aaeb88e0823bc35bca9"
   precision: "fp8"
 
 
 dynamo:
   install: true
-  version: 1.3.0.dev20260614
+  version: 1.3.0.dev20260710
 
 health_check:
   max_attempts: 720
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb200-fp8/8k1k/1p2d-dep4-tep4-8k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb200-fp8/8k1k/1p2d-dep4-tep4-8k1k.yaml
index 5bde9ac9be..64e712e0c8 100644
--- a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb200-fp8/8k1k/1p2d-dep4-tep4-8k1k.yaml
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb200-fp8/8k1k/1p2d-dep4-tep4-8k1k.yaml
@@ -6,13 +6,13 @@ name: "minimax-m3-vllm-disagg-gb200-1p2d-dep4-tep4-fp8-8k1k"
 
 model:
   path: "minimax-m3-mxfp8"
-  container: "vllm/vllm-openai:minimax-m3-perf-arm64-13.0.1-7a67223"
+  container: "vllm/vllm-openai:nightly-4080263bb2c5d10deac17aaeb88e0823bc35bca9"
   precision: "fp8"
 
 
 dynamo:
   install: true
-  version: 1.3.0.dev20260614
+  version: 1.3.0.dev20260710
 
 health_check:
   max_attempts: 720
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb200-fp8/8k1k/1p2d-dep4-tep8-8k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb200-fp8/8k1k/1p2d-dep4-tep8-8k1k.yaml
index 56ed02e5e8..8606cf5c38 100644
--- a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb200-fp8/8k1k/1p2d-dep4-tep8-8k1k.yaml
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb200-fp8/8k1k/1p2d-dep4-tep8-8k1k.yaml
@@ -6,13 +6,13 @@ name: "minimax-m3-vllm-disagg-gb200-1p2d-dep4-tep8-fp8-8k1k"
 
 model:
   path: "minimax-m3-mxfp8"
-  container: "vllm/vllm-openai:minimax-m3-perf-arm64-13.0.1-7a67223"
+  container: "vllm/vllm-openai:nightly-4080263bb2c5d10deac17aaeb88e0823bc35bca9"
   precision: "fp8"
 
 
 dynamo:
   install: true
-  version: 1.3.0.dev20260614
+  version: 1.3.0.dev20260710
 
 health_check:
   max_attempts: 720
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb200-fp8/8k1k/2p2d-dep4-dep8-8k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb200-fp8/8k1k/2p2d-dep4-dep8-8k1k.yaml
index 22ebfc7836..c5e9b185a7 100644
--- a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb200-fp8/8k1k/2p2d-dep4-dep8-8k1k.yaml
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb200-fp8/8k1k/2p2d-dep4-dep8-8k1k.yaml
@@ -6,13 +6,13 @@ name: "minimax-m3-vllm-disagg-gb200-2p2d-dep4-dep8-fp8-8k1k"
 
 model:
   path: "minimax-m3-mxfp8"
-  container: "vllm/vllm-openai:minimax-m3-perf-arm64-13.0.1-7a67223"
+  container: "vllm/vllm-openai:nightly-4080263bb2c5d10deac17aaeb88e0823bc35bca9"
   precision: "fp8"
 
 
 dynamo:
   install: true
-  version: 1.3.0.dev20260614
+  version: 1.3.0.dev20260710
 
 health_check:
   max_attempts: 720
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb200-fp8/8k1k/3p2d-dep4-dep8-8k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb200-fp8/8k1k/3p2d-dep4-dep8-8k1k.yaml
index ba1db206e4..ca89f94a72 100644
--- a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb200-fp8/8k1k/3p2d-dep4-dep8-8k1k.yaml
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb200-fp8/8k1k/3p2d-dep4-dep8-8k1k.yaml
@@ -6,13 +6,13 @@ name: "minimax-m3-vllm-disagg-gb200-3p2d-dep4-dep8-fp8-8k1k"
 
 model:
   path: "minimax-m3-mxfp8"
-  container: "vllm/vllm-openai:minimax-m3-perf-arm64-13.0.1-7a67223"
+  container: "vllm/vllm-openai:nightly-4080263bb2c5d10deac17aaeb88e0823bc35bca9"
   precision: "fp8"
 
 
 dynamo:
   install: true
-  version: 1.3.0.dev20260614
+  version: 1.3.0.dev20260710
 
 health_check:
   max_attempts: 720
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb200-fp8/8k1k/5p2d-dep4-dep8-8k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb200-fp8/8k1k/5p2d-dep4-dep8-8k1k.yaml
index a54ddcb226..df1ba1d8de 100644
--- a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb200-fp8/8k1k/5p2d-dep4-dep8-8k1k.yaml
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb200-fp8/8k1k/5p2d-dep4-dep8-8k1k.yaml
@@ -6,13 +6,13 @@ name: "minimax-m3-vllm-disagg-gb200-5p2d-dep4-dep8-fp8-8k1k"
 
 model:
   path: "minimax-m3-mxfp8"
-  container: "vllm/vllm-openai:minimax-m3-perf-arm64-13.0.1-7a67223"
+  container: "vllm/vllm-openai:nightly-4080263bb2c5d10deac17aaeb88e0823bc35bca9"
   precision: "fp8"
 
 
 dynamo:
   install: true
-  version: 1.3.0.dev20260614
+  version: 1.3.0.dev20260710
 
 health_check:
   max_attempts: 720
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb300-fp8/8k1k/1p1d-dep2-dep8-8k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb300-fp8/8k1k/1p1d-dep2-dep8-8k1k.yaml
index d990d661b9..ec72d22b08 100644
--- a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb300-fp8/8k1k/1p1d-dep2-dep8-8k1k.yaml
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb300-fp8/8k1k/1p1d-dep2-dep8-8k1k.yaml
@@ -6,13 +6,13 @@ name: "minimax-m3-vllm-disagg-gb300-1p1d-dep2-dep8-fp8-8k1k"
 
 model:
   path: "minimax-m3-mxfp8"
-  container: "vllm/vllm-openai:minimax-m3-perf-arm64-13.0.1-7a67223"
+  container: "vllm/vllm-openai:nightly-4080263bb2c5d10deac17aaeb88e0823bc35bca9"
   precision: "fp8"
 
 
 dynamo:
   install: true
-  version: 1.3.0.dev20260614
+  version: 1.3.0.dev20260710
 
 health_check:
   max_attempts: 720
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb300-fp8/8k1k/1p1d-dep2-tep8-8k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb300-fp8/8k1k/1p1d-dep2-tep8-8k1k.yaml
index d46133924c..ebc16ae605 100644
--- a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb300-fp8/8k1k/1p1d-dep2-tep8-8k1k.yaml
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb300-fp8/8k1k/1p1d-dep2-tep8-8k1k.yaml
@@ -6,13 +6,13 @@ name: "minimax-m3-vllm-disagg-gb300-1p1d-dep2-tep8-fp8-8k1k"
 
 model:
   path: "minimax-m3-mxfp8"
-  container: "vllm/vllm-openai:minimax-m3-perf-arm64-13.0.1-7a67223"
+  container: "vllm/vllm-openai:nightly-4080263bb2c5d10deac17aaeb88e0823bc35bca9"
   precision: "fp8"
 
 
 dynamo:
   install: true
-  version: 1.3.0.dev20260614
+  version: 1.3.0.dev20260710
 
 health_check:
   max_attempts: 720
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb300-fp8/8k1k/1p2d-dep2-tep8-8k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb300-fp8/8k1k/1p2d-dep2-tep8-8k1k.yaml
index e8c606e271..65a6f797a6 100644
--- a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb300-fp8/8k1k/1p2d-dep2-tep8-8k1k.yaml
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb300-fp8/8k1k/1p2d-dep2-tep8-8k1k.yaml
@@ -6,13 +6,13 @@ name: "minimax-m3-vllm-disagg-gb300-1p2d-dep2-tep8-fp8-8k1k"
 
 model:
   path: "minimax-m3-mxfp8"
-  container: "vllm/vllm-openai:minimax-m3-perf-arm64-13.0.1-7a67223"
+  container: "vllm/vllm-openai:nightly-4080263bb2c5d10deac17aaeb88e0823bc35bca9"
   precision: "fp8"
 
 
 dynamo:
   install: true
-  version: 1.3.0.dev20260614
+  version: 1.3.0.dev20260710
 
 health_check:
   max_attempts: 720
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb300-fp8/8k1k/2p1d-dep2-dep8-8k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb300-fp8/8k1k/2p1d-dep2-dep8-8k1k.yaml
index 02c3be14ac..fa51ba1792 100644
--- a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb300-fp8/8k1k/2p1d-dep2-dep8-8k1k.yaml
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb300-fp8/8k1k/2p1d-dep2-dep8-8k1k.yaml
@@ -6,13 +6,13 @@ name: "minimax-m3-vllm-disagg-gb300-2p1d-dep2-dep8-fp8-8k1k"
 
 model:
   path: "minimax-m3-mxfp8"
-  container: "vllm/vllm-openai:minimax-m3-perf-arm64-13.0.1-7a67223"
+  container: "vllm/vllm-openai:nightly-4080263bb2c5d10deac17aaeb88e0823bc35bca9"
   precision: "fp8"
 
 
 dynamo:
   install: true
-  version: 1.3.0.dev20260614
+  version: 1.3.0.dev20260710
 
 health_check:
   max_attempts: 720
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb300-fp8/8k1k/2p2d-dep2-tep8-8k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb300-fp8/8k1k/2p2d-dep2-tep8-8k1k.yaml
index 304650d6c1..79c7649689 100644
--- a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb300-fp8/8k1k/2p2d-dep2-tep8-8k1k.yaml
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb300-fp8/8k1k/2p2d-dep2-tep8-8k1k.yaml
@@ -6,13 +6,13 @@ name: "minimax-m3-vllm-disagg-gb300-2p2d-dep2-tep8-fp8-8k1k"
 
 model:
   path: "minimax-m3-mxfp8"
-  container: "vllm/vllm-openai:minimax-m3-perf-arm64-13.0.1-7a67223"
+  container: "vllm/vllm-openai:nightly-4080263bb2c5d10deac17aaeb88e0823bc35bca9"
   precision: "fp8"
 
 
 dynamo:
   install: true
-  version: 1.3.0.dev20260614
+  version: 1.3.0.dev20260710
 
 health_check:
   max_attempts: 720
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb300-fp8/8k1k/2p4d-dep2-tep4-8k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb300-fp8/8k1k/2p4d-dep2-tep4-8k1k.yaml
index efea8bfacf..fdb35c7fc6 100644
--- a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb300-fp8/8k1k/2p4d-dep2-tep4-8k1k.yaml
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb300-fp8/8k1k/2p4d-dep2-tep4-8k1k.yaml
@@ -6,13 +6,13 @@ name: "minimax-m3-vllm-disagg-gb300-2p4d-dep2-tep4-fp8-8k1k"
 
 model:
   path: "minimax-m3-mxfp8"
-  container: "vllm/vllm-openai:minimax-m3-perf-arm64-13.0.1-7a67223"
+  container: "vllm/vllm-openai:nightly-4080263bb2c5d10deac17aaeb88e0823bc35bca9"
   precision: "fp8"
 
 
 dynamo:
   install: true
-  version: 1.3.0.dev20260614
+  version: 1.3.0.dev20260710
 
 health_check:
   max_attempts: 720
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb300-fp8/8k1k/3p1d-dep2-dep8-8k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb300-fp8/8k1k/3p1d-dep2-dep8-8k1k.yaml
index 97e1ec88c8..8335b2f7bf 100644
--- a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb300-fp8/8k1k/3p1d-dep2-dep8-8k1k.yaml
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb300-fp8/8k1k/3p1d-dep2-dep8-8k1k.yaml
@@ -6,13 +6,13 @@ name: "minimax-m3-vllm-disagg-gb300-3p1d-dep2-dep8-fp8-8k1k"
 
 model:
   path: "minimax-m3-mxfp8"
-  container: "vllm/vllm-openai:minimax-m3-perf-arm64-13.0.1-7a67223"
+  container: "vllm/vllm-openai:nightly-4080263bb2c5d10deac17aaeb88e0823bc35bca9"
   precision: "fp8"
 
 
 dynamo:
   install: true
-  version: 1.3.0.dev20260614
+  version: 1.3.0.dev20260710
 
 health_check:
   max_attempts: 720
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb300-fp8/8k1k/3p2d-dep2-dep8-8k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb300-fp8/8k1k/3p2d-dep2-dep8-8k1k.yaml
index 745b2fad44..b52ebd3a2f 100644
--- a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb300-fp8/8k1k/3p2d-dep2-dep8-8k1k.yaml
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb300-fp8/8k1k/3p2d-dep2-dep8-8k1k.yaml
@@ -6,13 +6,13 @@ name: "minimax-m3-vllm-disagg-gb300-3p2d-dep2-dep8-fp8-8k1k"
 
 model:
   path: "minimax-m3-mxfp8"
-  container: "vllm/vllm-openai:minimax-m3-perf-arm64-13.0.1-7a67223"
+  container: "vllm/vllm-openai:nightly-4080263bb2c5d10deac17aaeb88e0823bc35bca9"
   precision: "fp8"
 
 
 dynamo:
   install: true
-  version: 1.3.0.dev20260614
+  version: 1.3.0.dev20260710
 
 health_check:
   max_attempts: 720
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb300-fp8/8k1k/6p1d-dep2-dep8-8k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb300-fp8/8k1k/6p1d-dep2-dep8-8k1k.yaml
index 9be5cc1772..94e7898237 100644
--- a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb300-fp8/8k1k/6p1d-dep2-dep8-8k1k.yaml
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb300-fp8/8k1k/6p1d-dep2-dep8-8k1k.yaml
@@ -6,13 +6,13 @@ name: "minimax-m3-vllm-disagg-gb300-6p1d-dep2-dep8-fp8-8k1k"
 
 model:
   path: "minimax-m3-mxfp8"
-  container: "vllm/vllm-openai:minimax-m3-perf-arm64-13.0.1-7a67223"
+  container: "vllm/vllm-openai:nightly-4080263bb2c5d10deac17aaeb88e0823bc35bca9"
   precision: "fp8"
 
 
 dynamo:
   install: true
-  version: 1.3.0.dev20260614
+  version: 1.3.0.dev20260710
 
 health_check:
   max_attempts: 720
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/1p1d-dep2-tp4-marlin-8k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/1p1d-dep2-tp4-marlin-8k1k.yaml
index 04aca65862..a51cfbd2a3 100644
--- a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/1p1d-dep2-tp4-marlin-8k1k.yaml
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/1p1d-dep2-tp4-marlin-8k1k.yaml
@@ -2,7 +2,7 @@ name: "minimax-m3-vllm-disagg-b300-1p1d-fp8-dep2-tp4-marlin-8k1k"
 
 model:
   path: "MiniMaxAI/MiniMax-M3-MXFP8"
-  container: "vllm/vllm-openai:minimax-m3-0618-x86_64-cu130"
+  container: "vllm/vllm-openai:nightly-4080263bb2c5d10deac17aaeb88e0823bc35bca9"
   precision: "fp8"
 
 resources:
@@ -17,7 +17,7 @@ resources:
 
 dynamo:
   install: true
-  version: 1.3.0.dev20260614
+  version: 1.3.0.dev20260710
 
 frontend:
   type: dynamo
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/1p2d-dep2-tep4-8k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/1p2d-dep2-tep4-8k1k.yaml
index e483108986..a962183696 100644
--- a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/1p2d-dep2-tep4-8k1k.yaml
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/1p2d-dep2-tep4-8k1k.yaml
@@ -2,7 +2,7 @@ name: "minimax-m3-vllm-disagg-b300-1p2d-fp8-dep2-tep4-8k1k"
 
 model:
   path: "MiniMaxAI/MiniMax-M3-MXFP8"
-  container: "vllm/vllm-openai:minimax-m3-0618-x86_64-cu130"
+  container: "vllm/vllm-openai:nightly-4080263bb2c5d10deac17aaeb88e0823bc35bca9"
   precision: "fp8"
 
 resources:
@@ -17,7 +17,7 @@ resources:
 
 dynamo:
   install: true
-  version: 1.3.0.dev20260614
+  version: 1.3.0.dev20260710
 
 frontend:
   type: dynamo
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/1p4d-dep2-tep4-8k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/1p4d-dep2-tep4-8k1k.yaml
index 30ac635a9a..a41361ae66 100644
--- a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/1p4d-dep2-tep4-8k1k.yaml
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/1p4d-dep2-tep4-8k1k.yaml
@@ -2,7 +2,7 @@ name: "minimax-m3-vllm-disagg-b300-1p4d-fp8-dep2-tep4-8k1k"
 
 model:
   path: "MiniMaxAI/MiniMax-M3-MXFP8"
-  container: "vllm/vllm-openai:minimax-m3-0618-x86_64-cu130"
+  container: "vllm/vllm-openai:nightly-4080263bb2c5d10deac17aaeb88e0823bc35bca9"
   precision: "fp8"
 
 resources:
@@ -17,7 +17,7 @@ resources:
 
 dynamo:
   install: true
-  version: 1.3.0.dev20260614
+  version: 1.3.0.dev20260710
 
 frontend:
   type: dynamo
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/1p4d-dep2-tep8-8k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/1p4d-dep2-tep8-8k1k.yaml
index 46af72e46c..c1a97a0c07 100644
--- a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/1p4d-dep2-tep8-8k1k.yaml
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/1p4d-dep2-tep8-8k1k.yaml
@@ -2,7 +2,7 @@ name: "minimax-m3-vllm-disagg-b300-1p4d-fp8-dep2-tep8-8k1k"
 
 model:
   path: "MiniMaxAI/MiniMax-M3-MXFP8"
-  container: "vllm/vllm-openai:minimax-m3-0618-x86_64-cu130"
+  container: "vllm/vllm-openai:nightly-4080263bb2c5d10deac17aaeb88e0823bc35bca9"
   precision: "fp8"
 
 resources:
@@ -17,7 +17,7 @@ resources:
 
 dynamo:
   install: true
-  version: 1.3.0.dev20260614
+  version: 1.3.0.dev20260710
 
 frontend:
   type: dynamo
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/2p2d-dep2-dep8-8k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/2p2d-dep2-dep8-8k1k.yaml
index b1558ae34f..873271d90a 100644
--- a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/2p2d-dep2-dep8-8k1k.yaml
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/2p2d-dep2-dep8-8k1k.yaml
@@ -2,7 +2,7 @@ name: "minimax-m3-vllm-disagg-b300-2p2d-fp8-dep2-dep8-8k1k"
 
 model:
   path: "MiniMaxAI/MiniMax-M3-MXFP8"
-  container: "vllm/vllm-openai:minimax-m3-0618-x86_64-cu130"
+  container: "vllm/vllm-openai:nightly-4080263bb2c5d10deac17aaeb88e0823bc35bca9"
   precision: "fp8"
 
 resources:
@@ -17,7 +17,7 @@ resources:
 
 dynamo:
   install: true
-  version: 1.3.0.dev20260614
+  version: 1.3.0.dev20260710
 
 frontend:
   type: dynamo
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/2p2d-dep2-tep8-8k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/2p2d-dep2-tep8-8k1k.yaml
index 46aaa045da..71fcc81d96 100644
--- a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/2p2d-dep2-tep8-8k1k.yaml
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/2p2d-dep2-tep8-8k1k.yaml
@@ -2,7 +2,7 @@ name: "minimax-m3-vllm-disagg-b300-2p2d-fp8-dep2-tep8-8k1k"
 
 model:
   path: "MiniMaxAI/MiniMax-M3-MXFP8"
-  container: "vllm/vllm-openai:minimax-m3-0618-x86_64-cu130"
+  container: "vllm/vllm-openai:nightly-4080263bb2c5d10deac17aaeb88e0823bc35bca9"
   precision: "fp8"
 
 resources:
@@ -17,7 +17,7 @@ resources:
 
 dynamo:
   install: true
-  version: 1.3.0.dev20260614
+  version: 1.3.0.dev20260710
 
 frontend:
   type: dynamo
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/4p2d-dep2-dep8-8k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/4p2d-dep2-dep8-8k1k.yaml
index 3756103eed..e36f12dfa5 100644
--- a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/4p2d-dep2-dep8-8k1k.yaml
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/4p2d-dep2-dep8-8k1k.yaml
@@ -2,7 +2,7 @@ name: "minimax-m3-vllm-disagg-b300-4p2d-fp8-dep2-dep8-8k1k"
 
 model:
   path: "MiniMaxAI/MiniMax-M3-MXFP8"
-  container: "vllm/vllm-openai:minimax-m3-0618-x86_64-cu130"
+  container: "vllm/vllm-openai:nightly-4080263bb2c5d10deac17aaeb88e0823bc35bca9"
   precision: "fp8"
 
 resources:
@@ -17,7 +17,7 @@ resources:
 
 dynamo:
   install: true
-  version: 1.3.0.dev20260614
+  version: 1.3.0.dev20260710
 
 frontend:
   type: dynamo
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/4p2d-dep2-tep4-8k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/4p2d-dep2-tep4-8k1k.yaml
index c9f29f7852..f8a6503096 100644
--- a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/4p2d-dep2-tep4-8k1k.yaml
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/4p2d-dep2-tep4-8k1k.yaml
@@ -2,7 +2,7 @@ name: "minimax-m3-vllm-disagg-b300-4p2d-fp8-dep2-tep4-8k1k"
 
 model:
   path: "MiniMaxAI/MiniMax-M3-MXFP8"
-  container: "vllm/vllm-openai:minimax-m3-0618-x86_64-cu130"
+  container: "vllm/vllm-openai:nightly-4080263bb2c5d10deac17aaeb88e0823bc35bca9"
   precision: "fp8"
 
 resources:
@@ -17,7 +17,7 @@ resources:
 
 dynamo:
   install: true
-  version: 1.3.0.dev20260614
+  version: 1.3.0.dev20260710
 
 frontend:
   type: dynamo
diff --git a/benchmarks/single_node/fixed_seq_len/minimaxm3_fp8_b200.sh b/benchmarks/single_node/fixed_seq_len/minimaxm3_fp8_b200.sh
index 7ac314a09a..1b4e70fd9f 100755
--- a/benchmarks/single_node/fixed_seq_len/minimaxm3_fp8_b200.sh
+++ b/benchmarks/single_node/fixed_seq_len/minimaxm3_fp8_b200.sh
@@ -22,39 +22,6 @@ check_env_vars \
     RANDOM_RANGE_RATIO \
     RESULT_FILENAME
 
-# The 0618 image keeps MiniMax M3 top-k indices in a persistent
-# [head_kv, max_tokens, topK] buffer for CUDA graphs. Slicing that buffer to
-# the actual prefill length is non-contiguous when TP leaves multiple local KV
-# heads, and the MSA CSR builder rejects it. Materialize the slice until the
-# image includes this fix.
-python3 - <<'PYEOF' || { echo "MiniMax M3 MSA contiguity patch failed" >&2; exit 1; }
-import importlib.util
-import pathlib
-
-spec = importlib.util.find_spec("vllm")
-if spec is None or not spec.submodule_search_locations:
-    raise RuntimeError("Could not locate the installed vllm package")
-
-target = (
-    pathlib.Path(next(iter(spec.submodule_search_locations)))
-    / "models"
-    / "minimax_m3"
-    / "nvidia"
-    / "sparse_attention_msa.py"
-)
-src = target.read_text()
-old = "            prefill_topk = topk[:, nd:num_tokens, :]\n"
-new = "            prefill_topk = topk[:, nd:num_tokens, :].contiguous()\n"
-
-if new in src:
-    print(f"[minimax-m3-msa-patch] already applied: {target}")
-elif src.count(old) == 1:
-    target.write_text(src.replace(old, new, 1))
-    print(f"[minimax-m3-msa-patch] patched: {target}")
-else:
-    raise RuntimeError(f"Expected exactly one patch anchor in {target}")
-PYEOF
-
 if [[ -n "$SLURM_JOB_ID" ]]; then
   echo "JOB $SLURM_JOB_ID running on $SLURMD_NODENAME"
 fi
@@ -100,7 +67,7 @@ $PARALLEL_ARGS \
 --language-model-only \
 --max-cudagraph-capture-size 2048 \
 --max-num-batched-tokens "$((ISL * 2 ))" \
---stream-interval 20 --no-enable-prefix-caching \
+--stream-interval 32 --no-enable-prefix-caching \
 --trust-remote-code > $SERVER_LOG 2>&1 &
 
 SERVER_PID=$!
diff --git a/benchmarks/single_node/fixed_seq_len/minimaxm3_fp8_b200_mtp.sh b/benchmarks/single_node/fixed_seq_len/minimaxm3_fp8_b200_mtp.sh
index 51147129dc..555741679a 100644
--- a/benchmarks/single_node/fixed_seq_len/minimaxm3_fp8_b200_mtp.sh
+++ b/benchmarks/single_node/fixed_seq_len/minimaxm3_fp8_b200_mtp.sh
@@ -3,15 +3,15 @@
 # MiniMax-M3 MXFP8 B200 single-node vLLM recipe with EAGLE3 speculative
 # decoding — the repo's spec-decoding=mtp variant of minimaxm3_fp8_b200.sh
 # (https://recipes.vllm.ai/MiniMaxAI/MiniMax-M3). Adds the
-# Inferact/MiniMax-M3-EAGLE3 draft head via --speculative-config with 3
-# speculative tokens. Everything else keeps the non-MTP serve shape:
+# Inferact/MiniMax-M3-EAGLE3-GQA draft head via --speculative-config with 1
+# speculative token. Everything else keeps the non-MTP serve shape:
 # --block-size 128 is mandatory (MSA sparse_block_size is 128; the default 16
 # misaligns sparse indexing), and --language-model-only skips the vision
 # encoder for the text-only benchmark. dp-attn=true maps to DP×EP (DEP);
 # ep>1 maps to TP+EP (TEP).
 #
-# The target uses the FlashInfer TRT-LLM attention path. The EAGLE3 drafter is
-# pinned separately to TRITON_ATTN.
+# The target uses the FlashInfer TRT-LLM attention path. The EAGLE3-GQA drafter
+# is pinned separately to FLASH_ATTN.
 
 source "$(dirname "$0")/../../benchmark_lib.sh"
 
@@ -27,40 +27,7 @@ check_env_vars \
     RANDOM_RANGE_RATIO \
     RESULT_FILENAME
 
-# The 0618 image keeps MiniMax M3 top-k indices in a persistent
-# [head_kv, max_tokens, topK] buffer for CUDA graphs. Slicing that buffer to
-# the actual prefill length is non-contiguous when TP leaves multiple local KV
-# heads, and the MSA CSR builder rejects it. Materialize the slice until the
-# image includes this fix.
-python3 - <<'PYEOF' || { echo "MiniMax M3 MSA contiguity patch failed" >&2; exit 1; }
-import importlib.util
-import pathlib
-
-spec = importlib.util.find_spec("vllm")
-if spec is None or not spec.submodule_search_locations:
-    raise RuntimeError("Could not locate the installed vllm package")
-
-target = (
-    pathlib.Path(next(iter(spec.submodule_search_locations)))
-    / "models"
-    / "minimax_m3"
-    / "nvidia"
-    / "sparse_attention_msa.py"
-)
-src = target.read_text()
-old = "            prefill_topk = topk[:, nd:num_tokens, :]\n"
-new = "            prefill_topk = topk[:, nd:num_tokens, :].contiguous()\n"
-
-if new in src:
-    print(f"[minimax-m3-msa-patch] already applied: {target}")
-elif src.count(old) == 1:
-    target.write_text(src.replace(old, new, 1))
-    print(f"[minimax-m3-msa-patch] patched: {target}")
-else:
-    raise RuntimeError(f"Expected exactly one patch anchor in {target}")
-PYEOF
-
-DRAFT_MODEL="Inferact/MiniMax-M3-EAGLE3"
+DRAFT_MODEL="Inferact/MiniMax-M3-EAGLE3-GQA"
 
 if [[ -n "$SLURM_JOB_ID" ]]; then
   echo "JOB $SLURM_JOB_ID running on $SLURMD_NODENAME"
@@ -100,8 +67,8 @@ else
   PARALLEL_ARGS="--tensor-parallel-size=$TP"
 fi
 
-# use 3 speculative tokens for all configs for now
-NUM_SPEC_TOKENS=3
+# use 1 speculative token for all configs for now
+NUM_SPEC_TOKENS=1
 
 if [ "${EVAL_ONLY}" = "true" ]; then
     setup_eval_context
@@ -116,14 +83,13 @@ $PARALLEL_ARGS \
 --gpu-memory-utilization 0.90 \
 --max-model-len $MAX_MODEL_LEN \
 --block-size 128 \
---attention-config '{"backend": "FLASHINFER", "use_trtllm_attention": true}' \
---attention-config.indexer_kv_dtype "fp8" \
+--attention-config '{"backend": "FLASHINFER", "use_trtllm_attention": true, "indexer_kv_dtype": "fp8"}' \
 --kv-cache-dtype fp8 \
 --language-model-only \
 --max-cudagraph-capture-size 2048 \
 --max-num-batched-tokens "$((ISL * 2 ))" \
---speculative-config "{\"method\": \"eagle3\", \"model\": \"$DRAFT_MODEL_PATH\", \"num_speculative_tokens\": $NUM_SPEC_TOKENS, \"attention_backend\": \"TRITON_ATTN\"}" \
---stream-interval 20 --no-enable-prefix-caching \
+--speculative-config "{\"method\": \"eagle3\", \"model\": \"$DRAFT_MODEL_PATH\", \"num_speculative_tokens\": $NUM_SPEC_TOKENS, \"attention_backend\": \"FLASH_ATTN\"}" \
+--stream-interval 32 --no-enable-prefix-caching \
 --trust-remote-code > $SERVER_LOG 2>&1 &
 
 SERVER_PID=$!
diff --git a/benchmarks/single_node/fixed_seq_len/minimaxm3_fp8_b300.sh b/benchmarks/single_node/fixed_seq_len/minimaxm3_fp8_b300.sh
index c0ee15f1f0..e37d30c930 100755
--- a/benchmarks/single_node/fixed_seq_len/minimaxm3_fp8_b300.sh
+++ b/benchmarks/single_node/fixed_seq_len/minimaxm3_fp8_b300.sh
@@ -23,39 +23,6 @@ check_env_vars \
     RANDOM_RANGE_RATIO \
     RESULT_FILENAME
 
-# The 0618 image keeps MiniMax M3 top-k indices in a persistent
-# [head_kv, max_tokens, topK] buffer for CUDA graphs. Slicing that buffer to
-# the actual prefill length is non-contiguous when TP leaves multiple local KV
-# heads, and the MSA CSR builder rejects it. Materialize the slice until the
-# image includes this fix.
-python3 - <<'PYEOF' || { echo "MiniMax M3 MSA contiguity patch failed" >&2; exit 1; }
-import importlib.util
-import pathlib
-
-spec = importlib.util.find_spec("vllm")
-if spec is None or not spec.submodule_search_locations:
-    raise RuntimeError("Could not locate the installed vllm package")
-
-target = (
-    pathlib.Path(next(iter(spec.submodule_search_locations)))
-    / "models"
-    / "minimax_m3"
-    / "nvidia"
-    / "sparse_attention_msa.py"
-)
-src = target.read_text()
-old = "            prefill_topk = topk[:, nd:num_tokens, :]\n"
-new = "            prefill_topk = topk[:, nd:num_tokens, :].contiguous()\n"
-
-if new in src:
-    print(f"[minimax-m3-msa-patch] already applied: {target}")
-elif src.count(old) == 1:
-    target.write_text(src.replace(old, new, 1))
-    print(f"[minimax-m3-msa-patch] patched: {target}")
-else:
-    raise RuntimeError(f"Expected exactly one patch anchor in {target}")
-PYEOF
-
 # `hf download` creates the target dir if missing and is itself idempotent.
 # When MODEL_PATH is unset (stand-alone runs), fall back to the HF_HUB_CACHE.
 # Either way, MODEL_PATH is what the server is launched with.
@@ -108,7 +75,7 @@ $PARALLEL_ARGS \
 --language-model-only \
 --max-cudagraph-capture-size 2048 \
 --max-num-batched-tokens "$((ISL * 2 ))" \
---stream-interval 20 --no-enable-prefix-caching \
+--stream-interval 32 --no-enable-prefix-caching \
 --trust-remote-code > $SERVER_LOG 2>&1 &
 
 SERVER_PID=$!
diff --git a/benchmarks/single_node/fixed_seq_len/minimaxm3_fp8_b300_mtp.sh b/benchmarks/single_node/fixed_seq_len/minimaxm3_fp8_b300_mtp.sh
index 01bf23eb6d..8bd8f78c7b 100644
--- a/benchmarks/single_node/fixed_seq_len/minimaxm3_fp8_b300_mtp.sh
+++ b/benchmarks/single_node/fixed_seq_len/minimaxm3_fp8_b300_mtp.sh
@@ -3,13 +3,13 @@
 # MiniMax-M3 MXFP8 B300 single-node vLLM recipe with EAGLE3 speculative
 # decoding — the repo's spec-decoding=mtp variant of minimaxm3_fp8_b300.sh
 # (https://recipes.vllm.ai/MiniMaxAI/MiniMax-M3). Adds the
-# Inferact/MiniMax-M3-EAGLE3 draft head via --speculative-config with 3
-# speculative tokens. Everything else keeps the non-MTP serve shape:
+# Inferact/MiniMax-M3-EAGLE3-GQA draft head via --speculative-config with 1
+# speculative token. Everything else keeps the non-MTP serve shape:
 # --block-size 128 is mandatory (MSA sparse/index cache); the benchmark is
 # text-only, so --language-model-only frees the vision encoder's VRAM.
 #
-# The target uses the FlashInfer TRT-LLM attention path. The EAGLE3 drafter is
-# pinned separately to TRITON_ATTN.
+# The target uses the FlashInfer TRT-LLM attention path. The EAGLE3-GQA drafter
+# is pinned separately to FLASH_ATTN.
 
 source "$(dirname "$0")/../../benchmark_lib.sh"
 
@@ -25,40 +25,7 @@ check_env_vars \
     RANDOM_RANGE_RATIO \
     RESULT_FILENAME
 
-# The 0618 image keeps MiniMax M3 top-k indices in a persistent
-# [head_kv, max_tokens, topK] buffer for CUDA graphs. Slicing that buffer to
-# the actual prefill length is non-contiguous when TP leaves multiple local KV
-# heads, and the MSA CSR builder rejects it. Materialize the slice until the
-# image includes this fix.
-python3 - <<'PYEOF' || { echo "MiniMax M3 MSA contiguity patch failed" >&2; exit 1; }
-import importlib.util
-import pathlib
-
-spec = importlib.util.find_spec("vllm")
-if spec is None or not spec.submodule_search_locations:
-    raise RuntimeError("Could not locate the installed vllm package")
-
-target = (
-    pathlib.Path(next(iter(spec.submodule_search_locations)))
-    / "models"
-    / "minimax_m3"
-    / "nvidia"
-    / "sparse_attention_msa.py"
-)
-src = target.read_text()
-old = "            prefill_topk = topk[:, nd:num_tokens, :]\n"
-new = "            prefill_topk = topk[:, nd:num_tokens, :].contiguous()\n"
-
-if new in src:
-    print(f"[minimax-m3-msa-patch] already applied: {target}")
-elif src.count(old) == 1:
-    target.write_text(src.replace(old, new, 1))
-    print(f"[minimax-m3-msa-patch] patched: {target}")
-else:
-    raise RuntimeError(f"Expected exactly one patch anchor in {target}")
-PYEOF
-
-DRAFT_MODEL="Inferact/MiniMax-M3-EAGLE3"
+DRAFT_MODEL="Inferact/MiniMax-M3-EAGLE3-GQA"
 
 # `hf download` creates the target dir if missing and is itself idempotent.
 # When MODEL_PATH is unset (stand-alone runs), fall back to the HF_HUB_CACHE.
@@ -102,8 +69,8 @@ else
   PARALLEL_ARGS="--tensor-parallel-size=$TP"
 fi
 
-# use 3 speculative tokens for all configs for now
-NUM_SPEC_TOKENS=3
+# use 1 speculative token for all configs for now
+NUM_SPEC_TOKENS=1
 
 if [ "${EVAL_ONLY}" = "true" ]; then
     setup_eval_context
@@ -118,14 +85,13 @@ $PARALLEL_ARGS \
 --gpu-memory-utilization 0.90 \
 --max-model-len $MAX_MODEL_LEN \
 --block-size 128 \
---attention-config '{"backend": "FLASHINFER", "use_trtllm_attention": true}' \
---attention-config.indexer_kv_dtype "fp8" \
+--attention-config '{"backend": "FLASHINFER", "use_trtllm_attention": true, "indexer_kv_dtype": "fp8"}' \
 --kv-cache-dtype fp8 \
 --language-model-only \
 --max-cudagraph-capture-size 2048 \
 --max-num-batched-tokens "$((ISL * 2 ))" \
---speculative-config "{\"method\": \"eagle3\", \"model\": \"$DRAFT_MODEL_PATH\", \"num_speculative_tokens\": $NUM_SPEC_TOKENS, \"attention_backend\": \"TRITON_ATTN\"}" \
---stream-interval 20 --no-enable-prefix-caching \
+--speculative-config "{\"method\": \"eagle3\", \"model\": \"$DRAFT_MODEL_PATH\", \"num_speculative_tokens\": $NUM_SPEC_TOKENS, \"attention_backend\": \"FLASH_ATTN\"}" \
+--stream-interval 32 --no-enable-prefix-caching \
 --trust-remote-code > $SERVER_LOG 2>&1 &
 
 SERVER_PID=$!
diff --git a/configs/nvidia-master.yaml b/configs/nvidia-master.yaml
index a388cad95d..cdb274f117 100644
--- a/configs/nvidia-master.yaml
+++ b/configs/nvidia-master.yaml
@@ -6666,7 +6666,7 @@ qwen3.5-fp8-h100-sglang-agentic:
 # baseline above. The image includes vLLM PR #46380, so no runtime patch is
 # needed.
 minimaxm3-fp8-b300-dynamo-vllm:
-  image: vllm/vllm-openai:minimax-m3-0618-x86_64-cu130
+  image: vllm/vllm-openai:nightly-4080263bb2c5d10deac17aaeb88e0823bc35bca9
   model: MiniMaxAI/MiniMax-M3-MXFP8
   model-prefix: minimaxm3
   runner: b300
@@ -7021,7 +7021,7 @@ minimaxm3-fp4-b300-dynamo-vllm-mtp:
 # All prefill DEP2 (TP1 DP2 EP, 2 GPU/worker). Decode: DEP4, TEP8, DEP8, TEP4.
 # 4 GPU/node (GB300 NVL72). kv-cache-dtype=fp8. srun_options mem=0 required.
 minimaxm3-fp8-gb300-dynamo-vllm:
-  image: vllm/vllm-openai:minimax-m3-perf-arm64-13.0.1-7a67223
+  image: vllm/vllm-openai:nightly-4080263bb2c5d10deac17aaeb88e0823bc35bca9
   model: MiniMaxAI/MiniMax-M3-MXFP8
   model-prefix: minimaxm3
   runner: gb300-nv
@@ -7195,7 +7195,7 @@ qwen3.5-fp4-b200-trt:
 # All prefill DEP4 (TP1 DP4 EP, 4 GPU/worker). Decode: Marlin, TEP8, DEP8, TEP4.
 # 4 GPU/node (GB200 NVL72). FLASHINFER attention with FP8 KV cache.
 minimaxm3-fp8-gb200-dynamo-vllm:
-  image: vllm/vllm-openai:minimax-m3-perf-arm64-13.0.1-7a67223
+  image: vllm/vllm-openai:nightly-4080263bb2c5d10deac17aaeb88e0823bc35bca9
   model: MiniMaxAI/MiniMax-M3-MXFP8
   model-prefix: minimaxm3
   runner: gb200
@@ -7328,7 +7328,7 @@ qwen3.5-fp4-b200-trt-mtp:
 # sparse/index cache alignment). Weights are NOT SRE-staged; b300 falls back
 # to writable /data/models (see launch_b300-nv.sh MODEL_PATH split).
 minimaxm3-fp8-b300-vllm:
-  image: vllm/vllm-openai:minimax-m3-0618-x86_64-cu130
+  image: vllm/vllm-openai:nightly-4080263bb2c5d10deac17aaeb88e0823bc35bca9
   model: MiniMaxAI/MiniMax-M3-MXFP8
   model-prefix: minimaxm3
   runner: b300
@@ -7409,7 +7409,7 @@ minimaxm3-fp4-b300-vllm-mtp:
 # /lustre/fsw/gharunners/models/MiniMax-M3-MXFP8 (pre-downloaded, see
 # launch_b200-dgxc.sh).
 minimaxm3-fp8-b200-vllm:
-  image: vllm/vllm-openai:minimax-m3-0618-x86_64-cu130
+  image: vllm/vllm-openai:nightly-4080263bb2c5d10deac17aaeb88e0823bc35bca9
   model: MiniMaxAI/MiniMax-M3-MXFP8
   model-prefix: minimaxm3
   runner: b200-dgxc
@@ -7461,7 +7461,7 @@ minimaxm3-fp4-b200-vllm:
 # precedent: spec decode pays off at low/mid concurrency while acceptance
 # dilutes in big batches, and the draft weights + draft KV shave headroom.
 minimaxm3-fp8-b200-vllm-mtp:
-  image: vllm/vllm-openai:minimax-m3-0618-x86_64-cu130
+  image: vllm/vllm-openai:nightly-4080263bb2c5d10deac17aaeb88e0823bc35bca9
   model: MiniMaxAI/MiniMax-M3-MXFP8
   model-prefix: minimaxm3
   runner: b200-dgxc
@@ -7517,7 +7517,7 @@ minimaxm3-fp4-b200-vllm-mtp:
 # big batches, and the draft weights + draft KV shave headroom — tp2-ep2 is
 # dropped entirely since its KV headroom was already thin without a draft.
 minimaxm3-fp8-b300-vllm-mtp:
-  image: vllm/vllm-openai:minimax-m3-0618-x86_64-cu130
+  image: vllm/vllm-openai:nightly-4080263bb2c5d10deac17aaeb88e0823bc35bca9
   model: MiniMaxAI/MiniMax-M3-MXFP8
   model-prefix: minimaxm3
   runner: b300
diff --git a/perf-changelog.yaml b/perf-changelog.yaml
index 975c917743..4e01656723 100644
--- a/perf-changelog.yaml
+++ b/perf-changelog.yaml
@@ -5066,3 +5066,16 @@
   description:
     - "Bump SGLang container image from lmsysorg/sglang:v0.5.12-cu130 to lmsysorg/sglang:v0.5.15.post1-cu130 (https://github.com/sgl-project/sglang/releases/tag/v0.5.15.post1)"
   pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/2313
+
+- config-keys:
+    - minimaxm3-fp8-b300-vllm
+    - minimaxm3-fp8-b300-vllm-mtp
+    - minimaxm3-fp8-b200-vllm
+    - minimaxm3-fp8-b200-vllm-mtp
+    - minimaxm3-fp8-b300-dynamo-vllm
+    - minimaxm3-fp8-gb300-dynamo-vllm
+    - minimaxm3-fp8-gb200-dynamo-vllm
+  description:
+    - "Bump image to vllm/vllm-openai:nightly-4080263bb2c5d10deac17aaeb88e0823bc35bca9 and dynamo to 1.3.0.dev20260710"
+  pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/2334
+
diff --git a/runners/launch_b300-nv.sh b/runners/launch_b300-nv.sh
index 63c0916962..57e8bfd43a 100644
--- a/runners/launch_b300-nv.sh
+++ b/runners/launch_b300-nv.sh
@@ -94,9 +94,6 @@ elif [[ $FRAMEWORK == "dynamo-vllm" && $MODEL_PREFIX == "minimaxm3" && ( $PRECIS
     git checkout sa-submission-q2-2026
     mkdir -p recipes/vllm/minimax-m3
     cp -rT "$GITHUB_WORKSPACE/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3" recipes/vllm/minimax-m3
-    if [[ $PRECISION == "fp8" ]]; then
-        SRTCTL_SETUP_SCRIPT="minimax-m3-vllm-fixes.sh"
-    fi
     # NVIDIA/srt-slurm#38
     git show 22d46ba9971615016d2339c9ffbc7b4597accfad --format= -- src/srtctl/core/ip_utils/get_node_ip.sh | git apply - || exit 1
     if [[ -n "$SRTCTL_SETUP_SCRIPT" ]]; then
diff --git a/runners/launch_gb200-nv.sh b/runners/launch_gb200-nv.sh
index a5f917da09..169014e86c 100755
--- a/runners/launch_gb200-nv.sh
+++ b/runners/launch_gb200-nv.sh
@@ -347,10 +347,6 @@ elif [[ $FRAMEWORK == "dynamo-vllm" && $MODEL_PREFIX == "minimaxm3" && $PRECISIO
     git checkout sa-submission-q2-2026 || exit 1
     mkdir -p recipes/vllm/minimax-m3-gb200-fp8 || exit 1
     cp -rT "$GITHUB_WORKSPACE/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb200-fp8" recipes/vllm/minimax-m3-gb200-fp8 || exit 1
-    SRTCTL_SETUP_SCRIPT="minimax-m3-gb200-vllm-fixes.sh"
-    cp \
-        "$GITHUB_WORKSPACE/benchmarks/multi_node/srt-slurm-recipes/configs/$SRTCTL_SETUP_SCRIPT" \
-        "configs/$SRTCTL_SETUP_SCRIPT" || exit 1
 elif [[ $FRAMEWORK == "dynamo-vllm" && $MODEL_PREFIX == "kimik2.5" && $PRECISION == "fp4" ]]; then
     git clone https://github.com/NVIDIA/srt-slurm.git "$SRT_REPO_DIR" || exit 1
     cd "$SRT_REPO_DIR" || exit 1
diff --git a/runners/launch_gb300-nv.sh b/runners/launch_gb300-nv.sh
index d450cdb86d..9e2f994e27 100644
--- a/runners/launch_gb300-nv.sh
+++ b/runners/launch_gb300-nv.sh
@@ -234,10 +234,6 @@ elif [[ $FRAMEWORK == "dynamo-vllm" && $MODEL_PREFIX == "minimaxm3" ]]; then
     git checkout sa-submission-q2-2026
     mkdir -p recipes/vllm/minimax-m3-gb300-fp8
     cp -rT "$GITHUB_WORKSPACE/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3-gb300-fp8" recipes/vllm/minimax-m3-gb300-fp8
-    SRTCTL_SETUP_SCRIPT="minimax-m3-gb300-vllm-fixes.sh"
-    cp \
-        "$GITHUB_WORKSPACE/benchmarks/multi_node/srt-slurm-recipes/configs/$SRTCTL_SETUP_SCRIPT" \
-        "configs/$SRTCTL_SETUP_SCRIPT"
 elif [[ $FRAMEWORK == "dynamo-vllm" && $MODEL_PREFIX == "kimik2.5" && $PRECISION == "fp4" ]]; then
     git clone https://github.com/NVIDIA/srt-slurm.git "$SRT_REPO_DIR"
     cd "$SRT_REPO_DIR"