From 1be41e9036e19fc620d39e8f7b22572fa36751a9 Mon Sep 17 00:00:00 2001
From: Mohammad Miadh Angkad <176301910+mmangkad@users.noreply.github.com>
Date: Sat, 14 Feb 2026 07:43:33 +0800
Subject: [PATCH] [FlashInfer] Bump FlashInfer version from 0.6.2 to 0.6.3
 (#18448)

---
 .github/workflows/release-docker-cu13-framework.yml | 4 ++--
 docker/Dockerfile                                   | 2 +-
 python/pyproject.toml                               | 4 ++--
 python/sglang/srt/entrypoints/engine.py             | 2 +-
 python/sglang/srt/server_args.py                    | 4 ++--
 python/sglang/srt/utils/common.py                   | 2 +-
 scripts/ci/cuda/ci_install_dependency.sh            | 2 +-
 7 files changed, 10 insertions(+), 10 deletions(-)

diff --git a/.github/workflows/release-docker-cu13-framework.yml b/.github/workflows/release-docker-cu13-framework.yml
index 8674c28f8..a366b0091 100644
--- a/.github/workflows/release-docker-cu13-framework.yml
+++ b/.github/workflows/release-docker-cu13-framework.yml
@@ -10,9 +10,9 @@ on:
         description: "Version to build (without v prefix, e.g., 0.5.8)"
         required: true
       flashinfer_version:
-        description: "FlashInfer version (default: 0.6.1)"
+        description: "FlashInfer version (default: 0.6.3)"
         required: false
-        default: "0.6.1"
+        default: "0.6.3"
 
 jobs:
   publish-x86:
diff --git a/docker/Dockerfile b/docker/Dockerfile
index 234374b49..5723e8db3 100644
--- a/docker/Dockerfile
+++ b/docker/Dockerfile
@@ -19,7 +19,7 @@ ARG PIP_DEFAULT_INDEX
 ARG UBUNTU_MIRROR
 ARG GITHUB_ARTIFACTORY=github.com
 ARG INSTALL_FLASHINFER_JIT_CACHE=0
-ARG FLASHINFER_VERSION=0.6.2
+ARG FLASHINFER_VERSION=0.6.3
 ARG MOONCAKE_VERSION=0.3.9
 #if need other arg please add in MOONCAKE_COMPILE_ARG
 ARG MOONCAKE_COMPILE_ARG="-DUSE_HTTP=ON -DUSE_MNNVL=ON -DUSE_CUDA=ON -DWITH_EP=ON"
diff --git a/python/pyproject.toml b/python/pyproject.toml
index bb5d1d155..d8beb00df 100755
--- a/python/pyproject.toml
+++ b/python/pyproject.toml
@@ -28,8 +28,8 @@ dependencies = [
   "datasets",
   "einops",
   "fastapi",
-  "flashinfer_python==0.6.2", # keep it aligned with jit-cache version in Dockerfile
-  "flashinfer_cubin==0.6.2",
+  "flashinfer_python==0.6.3", # keep it aligned with jit-cache version in Dockerfile
+  "flashinfer_cubin==0.6.3",
   "gguf",
   "hf_transfer",
   "huggingface_hub",
diff --git a/python/sglang/srt/entrypoints/engine.py b/python/sglang/srt/entrypoints/engine.py
index b701349a2..b5324d8b5 100644
--- a/python/sglang/srt/entrypoints/engine.py
+++ b/python/sglang/srt/entrypoints/engine.py
@@ -841,7 +841,7 @@ def _set_envs_and_config(server_args: ServerArgs):
         if server_args.attention_backend == "flashinfer":
             assert_pkg_version(
                 "flashinfer_python",
-                "0.6.2",
+                "0.6.3",
                 "Please uninstall the old version and "
                 "reinstall the latest version by following the instructions "
                 "at https://docs.flashinfer.ai/installation.html.",
diff --git a/python/sglang/srt/server_args.py b/python/sglang/srt/server_args.py
index ebf312c17..79e0c51c2 100644
--- a/python/sglang/srt/server_args.py
+++ b/python/sglang/srt/server_args.py
@@ -1616,8 +1616,8 @@ class ServerArgs:
                     and self.moe_a2a_backend == "none"
                     and self.moe_runner_backend == "auto"
                 ):
-                    # Only enable flashinfer_trtllm if flashinfer-python version is >= 0.6.2
-                    if check_pkg_version_at_least("flashinfer-python", "0.6.2"):
+                    # Only enable flashinfer_trtllm if flashinfer-python version is >= 0.6.3
+                    if check_pkg_version_at_least("flashinfer-python", "0.6.3"):
                         self.moe_runner_backend = "flashinfer_trtllm"
                         logger.info(
                             "Use flashinfer_trtllm as MoE runner backend on sm100 for Glm4MoeForCausalLM"
diff --git a/python/sglang/srt/utils/common.py b/python/sglang/srt/utils/common.py
index 3bae1b955..edc073e72 100644
--- a/python/sglang/srt/utils/common.py
+++ b/python/sglang/srt/utils/common.py
@@ -1113,7 +1113,7 @@ def check_pkg_version_at_least(pkg: str, min_version: str) -> bool:
 
     Args:
         pkg: Package name (distribution name, e.g., "flashinfer-python")
-        min_version: Minimum version required (e.g., "0.6.2")
+        min_version: Minimum version required (e.g., "0.6.3")
 
     Returns:
         True if package is installed and version >= min_version, False otherwise
diff --git a/scripts/ci/cuda/ci_install_dependency.sh b/scripts/ci/cuda/ci_install_dependency.sh
index 4d8d20864..fd0349c47 100755
--- a/scripts/ci/cuda/ci_install_dependency.sh
+++ b/scripts/ci/cuda/ci_install_dependency.sh
@@ -5,7 +5,7 @@ set -euxo pipefail
 # Set up environment variables
 IS_BLACKWELL=${IS_BLACKWELL:-0}
 CU_VERSION="cu129"
-FLASHINFER_VERSION=0.6.2
+FLASHINFER_VERSION=0.6.3
 OPTIONAL_DEPS="${1:-}"
 
 # Detect system architecture