diff --git a/extensions/inference-cortex-extension/download.bat b/extensions/inference-cortex-extension/download.bat
index 0b13ee872..e19786971 100644
--- a/extensions/inference-cortex-extension/download.bat
+++ b/extensions/inference-cortex-extension/download.bat
@@ -3,8 +3,9 @@ set BIN_PATH=./bin
 set /p CORTEX_VERSION=<./bin/version.txt
 
 @REM Download cortex.llamacpp binaries
-set VERSION=v0.1.25
-set DOWNLOAD_URL=https://github.com/janhq/cortex.llamacpp/releases/download/%VERSION%/cortex.llamacpp-0.1.34-windows-amd64
+set VERSION=v0.1.35
+set DOWNLOAD_URL=https://github.com/janhq/cortex.llamacpp/releases/download/%VERSION%/cortex.llamacpp-0.1.35-windows-amd64
+set CUDA_DOWNLOAD_URL=https://github.com/janhq/cortex.llamacpp/releases/download/%VERSION%
 set SUBFOLDERS=win-cuda-12-0 win-cuda-11-7 win-noavx win-avx win-avx2 win-avx512 win-vulkan
 
 call .\node_modules\.bin\download -e --strip 1 -o %BIN_PATH% https://github.com/janhq/cortex/releases/download/v%CORTEX_VERSION%/cortex-%CORTEX_VERSION%-windows-amd64.tar.gz
@@ -17,8 +18,8 @@ call .\node_modules\.bin\download %DOWNLOAD_URL%-avx.tar.gz -e --strip 1 -o %BIN
 call .\node_modules\.bin\download %DOWNLOAD_URL%-avx2.tar.gz -e --strip 1 -o %BIN_PATH%/avx2/engines/cortex.llamacpp
 call .\node_modules\.bin\download %DOWNLOAD_URL%-avx512.tar.gz -e --strip 1 -o %BIN_PATH%/avx512/engines/cortex.llamacpp
 call .\node_modules\.bin\download %DOWNLOAD_URL%-vulkan.tar.gz -e --strip 1 -o %BIN_PATH%/vulkan/engines/cortex.llamacpp
-call .\node_modules\.bin\download %DOWNLOAD_URL%-cuda-12-0-windows-amd64.tar.gz -e --strip 1 -o %BIN_PATH%
-call .\node_modules\.bin\download %DOWNLOAD_URL%-cuda-11-7-windows-amd64.tar.gz -e --strip 1 -o %BIN_PATH%
+call .\node_modules\.bin\download %CUDA_DOWNLOAD_URL%-cuda-12-0-windows-amd64.tar.gz -e --strip 1 -o %BIN_PATH%
+call .\node_modules\.bin\download %CUDA_DOWNLOAD_URL%-cuda-11-7-windows-amd64.tar.gz -e --strip 1 -o %BIN_PATH%
 
 
 @REM Loop through each folder and move DLLs (excluding engine.dll)
diff --git a/extensions/inference-cortex-extension/download.sh b/extensions/inference-cortex-extension/download.sh
index d04f0482d..9426bc872 100755
--- a/extensions/inference-cortex-extension/download.sh
+++ b/extensions/inference-cortex-extension/download.sh
@@ -3,7 +3,8 @@
 # Read CORTEX_VERSION
 CORTEX_VERSION=$(cat ./bin/version.txt)
 CORTEX_RELEASE_URL="https://github.com/janhq/cortex/releases/download"
-ENGINE_DOWNLOAD_URL="https://github.com/janhq/cortex.llamacpp/releases/download/v0.1.34/cortex.llamacpp-0.1.34"
+ENGINE_DOWNLOAD_URL="https://github.com/janhq/cortex.llamacpp/releases/download/v0.1.35/cortex.llamacpp-0.1.35"
+CUDA_DOWNLOAD_URL="https://github.com/janhq/cortex.llamacpp/releases/download/v0.1.35"
 # Detect platform
 OS_TYPE=$(uname)
 
@@ -22,8 +23,8 @@ if [ "$OS_TYPE" == "Linux" ]; then
     download "${ENGINE_DOWNLOAD_URL}-linux-amd64-noavx-cuda-12-0.tar.gz" -e --strip 1 -o "./bin/noavx-cuda-12-0/engines/cortex.llamacpp" 1
     download "${ENGINE_DOWNLOAD_URL}-linux-amd64-noavx-cuda-11-7.tar.gz" -e --strip 1 -o "./bin/noavx-cuda-11-7/engines/cortex.llamacpp" 1
     download "${ENGINE_DOWNLOAD_URL}-linux-amd64-vulkan.tar.gz" -e --strip 1 -o "./bin/vulkan/engines/cortex.llamacpp" 1
-    download "${ENGINE_DOWNLOAD_URL}-cuda-12-0-linux-amd64.tar.gz" -e --strip 1 -o "./bin" 1
-    download "${ENGINE_DOWNLOAD_URL}-cuda-11-7-linux-amd64.tar.gz" -e --strip 1 -o "./bin" 1
+    download "${CUDA_DOWNLOAD_URL}-cuda-12-0-linux-amd64.tar.gz" -e --strip 1 -o "./bin" 1
+    download "${CUDA_DOWNLOAD_URL}-cuda-11-7-linux-amd64.tar.gz" -e --strip 1 -o "./bin" 1
 
 elif [ "$OS_TYPE" == "Darwin" ]; then
     # macOS downloads
diff --git a/extensions/inference-cortex-extension/src/node/execute.test.ts b/extensions/inference-cortex-extension/src/node/execute.test.ts
index 3c0b32df5..622eb38af 100644
--- a/extensions/inference-cortex-extension/src/node/execute.test.ts
+++ b/extensions/inference-cortex-extension/src/node/execute.test.ts
@@ -46,11 +46,11 @@ describe('test executable cortex file', () => {
     })
     expect(executableCortexFile(testSettings)).toEqual(
       expect.objectContaining({
-        enginePath: expect.stringContaining(`/bin/arm64`),
-        binPath: expect.stringContaining(`/bin`),
+        enginePath: expect.stringContaining(`arm64`),
+        binPath: expect.stringContaining(`bin`),
         executablePath:
           originalPlatform === 'darwin'
-            ? expect.stringContaining(`/cortex-server`)
+            ? expect.stringContaining(`cortex-server`)
             : expect.anything(),
         cudaVisibleDevices: '',
         vkVisibleDevices: '',
@@ -61,11 +61,11 @@ describe('test executable cortex file', () => {
     })
     expect(executableCortexFile(testSettings)).toEqual(
       expect.objectContaining({
-        enginePath: expect.stringContaining(`/bin/x64`),
-        binPath: expect.stringContaining(`/bin`),
+        enginePath: expect.stringContaining(`x64`),
+        binPath: expect.stringContaining(`bin`),
         executablePath:
           originalPlatform === 'darwin'
-            ? expect.stringContaining(`/cortex-server`)
+            ? expect.stringContaining(`cortex-server`)
             : expect.anything(),
         cudaVisibleDevices: '',
         vkVisibleDevices: '',
@@ -84,9 +84,9 @@ describe('test executable cortex file', () => {
     mockCpuInfo.mockReturnValue(['avx'])
     expect(executableCortexFile(settings)).toEqual(
       expect.objectContaining({
-        enginePath: expect.stringContaining(`/bin/avx`),
-        binPath: expect.stringContaining(`/bin`),
-        executablePath: expect.stringContaining(`/cortex-server.exe`),
+        enginePath: expect.stringContaining(`avx`),
+        binPath: expect.stringContaining(`bin`),
+        executablePath: expect.stringContaining(`cortex-server.exe`),
         cudaVisibleDevices: '',
         vkVisibleDevices: '',
       })
@@ -117,11 +117,12 @@ describe('test executable cortex file', () => {
         },
       ],
     }
+    mockCpuInfo.mockReturnValue(['avx2'])
     expect(executableCortexFile(settings)).toEqual(
       expect.objectContaining({
-        enginePath: expect.stringContaining(`cuda-11-7`),
-        binPath: expect.stringContaining(`/bin`),
-        executablePath: expect.stringContaining(`/cortex-server.exe`),
+        enginePath: expect.stringContaining(`avx2-cuda-11-7`),
+        binPath: expect.stringContaining(`bin`),
+        executablePath: expect.stringContaining(`cortex-server.exe`),
         cudaVisibleDevices: '0',
         vkVisibleDevices: '0',
       })
@@ -152,11 +153,12 @@ describe('test executable cortex file', () => {
         },
       ],
     }
+    mockCpuInfo.mockReturnValue(['noavx'])
     expect(executableCortexFile(settings)).toEqual(
       expect.objectContaining({
-        enginePath: expect.stringContaining(`cuda-12-0`),
-        binPath: expect.stringContaining(`/bin`),
-        executablePath: expect.stringContaining(`/cortex-server.exe`),
+        enginePath: expect.stringContaining(`noavx-cuda-12-0`),
+        binPath: expect.stringContaining(`bin`),
+        executablePath: expect.stringContaining(`cortex-server.exe`),
         cudaVisibleDevices: '0',
         vkVisibleDevices: '0',
       })
@@ -175,7 +177,7 @@ describe('test executable cortex file', () => {
     expect(executableCortexFile(settings)).toEqual(
       expect.objectContaining({
         enginePath: expect.stringContaining(`noavx`),
-        executablePath: expect.stringContaining(`/cortex-server`),
+        executablePath: expect.stringContaining(`cortex-server`),
         cudaVisibleDevices: '',
         vkVisibleDevices: '',
       })
@@ -209,8 +211,8 @@ describe('test executable cortex file', () => {
     expect(executableCortexFile(settings)).toEqual(
       expect.objectContaining({
         enginePath: expect.stringContaining(`cuda-11-7`),
-        binPath: expect.stringContaining(`/bin`),
-        executablePath: expect.stringContaining(`/cortex-server`),
+        binPath: expect.stringContaining(`bin`),
+        executablePath: expect.stringContaining(`cortex-server`),
         cudaVisibleDevices: '0',
         vkVisibleDevices: '0',
       })
@@ -244,8 +246,8 @@ describe('test executable cortex file', () => {
     expect(executableCortexFile(settings)).toEqual(
       expect.objectContaining({
         enginePath: expect.stringContaining(`cuda-12-0`),
-        binPath: expect.stringContaining(`/bin`),
-        executablePath: expect.stringContaining(`/cortex-server`),
+        binPath: expect.stringContaining(`bin`),
+        executablePath: expect.stringContaining(`cortex-server`),
         cudaVisibleDevices: '0',
         vkVisibleDevices: '0',
       })
@@ -269,8 +271,8 @@ describe('test executable cortex file', () => {
       expect(executableCortexFile(settings)).toEqual(
         expect.objectContaining({
           enginePath: expect.stringContaining(instruction),
-          binPath: expect.stringContaining(`/bin`),
-          executablePath: expect.stringContaining(`/cortex-server`),
+          binPath: expect.stringContaining(`bin`),
+          executablePath: expect.stringContaining(`cortex-server`),
 
           cudaVisibleDevices: '',
           vkVisibleDevices: '',
@@ -293,8 +295,8 @@ describe('test executable cortex file', () => {
       expect(executableCortexFile(settings)).toEqual(
         expect.objectContaining({
           enginePath: expect.stringContaining(instruction),
-          binPath: expect.stringContaining(`/bin`),
-          executablePath: expect.stringContaining(`/cortex-server.exe`),
+          binPath: expect.stringContaining(`bin`),
+          executablePath: expect.stringContaining(`cortex-server.exe`),
           cudaVisibleDevices: '',
           vkVisibleDevices: '',
         })
@@ -333,8 +335,8 @@ describe('test executable cortex file', () => {
       expect(executableCortexFile(settings)).toEqual(
         expect.objectContaining({
           enginePath: expect.stringContaining(`cuda-12-0`),
-          binPath: expect.stringContaining(`/bin`),
-          executablePath: expect.stringContaining(`/cortex-server.exe`),
+          binPath: expect.stringContaining(`bin`),
+          executablePath: expect.stringContaining(`cortex-server.exe`),
           cudaVisibleDevices: '0',
           vkVisibleDevices: '0',
         })
@@ -373,8 +375,8 @@ describe('test executable cortex file', () => {
       expect(executableCortexFile(settings)).toEqual(
         expect.objectContaining({
           enginePath: expect.stringContaining(`cuda-12-0`),
-          binPath: expect.stringContaining(`/bin`),
-          executablePath: expect.stringContaining(`/cortex-server`),
+          binPath: expect.stringContaining(`bin`),
+          executablePath: expect.stringContaining(`cortex-server`),
           cudaVisibleDevices: '0',
           vkVisibleDevices: '0',
         })
@@ -414,8 +416,8 @@ describe('test executable cortex file', () => {
       expect(executableCortexFile(settings)).toEqual(
         expect.objectContaining({
           enginePath: expect.stringContaining(`vulkan`),
-          binPath: expect.stringContaining(`/bin`),
-          executablePath: expect.stringContaining(`/cortex-server`),
+          binPath: expect.stringContaining(`bin`),
+          executablePath: expect.stringContaining(`cortex-server`),
           cudaVisibleDevices: '0',
           vkVisibleDevices: '0',
         })
@@ -441,10 +443,10 @@ describe('test executable cortex file', () => {
       expect(executableCortexFile(settings)).toEqual(
         expect.objectContaining({
           enginePath: expect.stringContaining(`x64`),
-          binPath: expect.stringContaining(`/bin`),
+          binPath: expect.stringContaining(`bin`),
           executablePath:
             originalPlatform === 'darwin'
-              ? expect.stringContaining(`/cortex-server`)
+              ? expect.stringContaining(`cortex-server`)
               : expect.anything(),
           cudaVisibleDevices: '',
           vkVisibleDevices: '',
diff --git a/extensions/inference-cortex-extension/src/node/execute.ts b/extensions/inference-cortex-extension/src/node/execute.ts
index 18d840fdd..b8e768bb9 100644
--- a/extensions/inference-cortex-extension/src/node/execute.ts
+++ b/extensions/inference-cortex-extension/src/node/execute.ts
@@ -92,10 +92,9 @@ export const executableCortexFile = (
       ? os()
       : [
           gpuRunMode(gpuSetting) !== 'cuda' ||
-          cpuInstruction === 'avx' ||
-          cpuInstruction === 'noavx'
+          cpuInstruction === 'avx2'
             ? cpuInstruction
-            : '',
+            : 'noavx',
           gpuRunMode(gpuSetting),
           cudaVersion(gpuSetting),
         ]