diff --git a/install.sh b/install.sh
index cc4d5a2..5cfaef2 100755
--- a/install.sh
+++ b/install.sh
@@ -134,6 +134,12 @@ main() {
   cd -
   rm -rf vllm-$vllm_v*
 
+  # Upgrade transformers beyond vllm's <5 pin.
+  # mlx-lm 0.30+ and mlx-vlm 0.3.10+ require transformers>=5.0.0 for newer
+  # model architectures (Qwen3.5, Nemotron, etc.). vllm works fine with v5 —
+  # upstream is tracking the official upgrade in vllm-project/vllm#30566.
+  uv pip install 'transformers>=5.0.0'
+
   if [[ -n "$local_lib" && -f "$local_lib" ]]; then
     uv pip install .
   else
diff --git a/pyproject.toml b/pyproject.toml
index 9a5cd7c..b3a2f7c 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -32,7 +32,7 @@ dependencies = [
     "mlx-lm>=0.28.4; platform_system == 'Darwin' and platform_machine == 'arm64'",
     "mlx-vlm>=0.3.0; platform_system == 'Darwin' and platform_machine == 'arm64'",  # Vision-language model support
     # Model loading and weights
-    "transformers>=4.40.0",
+    "transformers>=5.0.0",
     "accelerate>=0.26.0",
     "safetensors>=0.4.0",
     # Native Metal extension JIT build