diff --git a/benchmarks/kernels/deepgemm/README.md b/benchmarks/kernels/deepgemm/README.md
index 41e68e047be8..a28c6956be0e 100644
--- a/benchmarks/kernels/deepgemm/README.md
+++ b/benchmarks/kernels/deepgemm/README.md
@@ -2,7 +2,7 @@
 
 This directory includes benchmarks between DeepSeek's DeepGEMM block fp8 kernels against vLLM's existing triton and CUTLASS-based kernels.
 
-Currently this just includes dense GEMMs and only works on Hopper GPUs.
+Currently, this just includes dense GEMMs and only works on Hopper GPUs.
 
 ## Setup
 
diff --git a/vllm/config/vllm.py b/vllm/config/vllm.py
index d64e315b4fe3..8a3599416bc7 100644
--- a/vllm/config/vllm.py
+++ b/vllm/config/vllm.py
@@ -96,7 +96,7 @@ class VllmConfig:
     """`torch.compile` and cudagraph capture configuration for the model.
 
     As a shorthand, one can append compilation arguments via 
-    -0.parameter=arguement such as `-O.mode=3` (same as `-O='{"mode":3}'`).
+    -0.parameter=argument such as `-O.mode=3` (same as `-O='{"mode":3}'`).
 
     You can specify the full compilation config like so:
     `{"mode": 3, "cudagraph_capture_sizes": [1, 2, 4, 8]}`
diff --git a/vllm/forward_context.py b/vllm/forward_context.py
index 25fb7181a8f2..7cb490e391ab 100644
--- a/vllm/forward_context.py
+++ b/vllm/forward_context.py
@@ -153,7 +153,7 @@ class DPMetadata:
     @contextmanager
     def sp_local_sizes(self, sequence_parallel_size: int):
         """
-        Context mamager for setting self.local_sizes. Same as self.chunked_sizes
+        Context manager for setting self.local_sizes. Same as self.chunked_sizes
         but without any chunking.
         """
         self.local_sizes = _compute_sp_num_tokens(
diff --git a/vllm/v1/worker/gpu_input_batch.py b/vllm/v1/worker/gpu_input_batch.py
index d6fef450c028..4a2818ab1bfd 100644
--- a/vllm/v1/worker/gpu_input_batch.py
+++ b/vllm/v1/worker/gpu_input_batch.py
@@ -525,7 +525,7 @@ class InputBatch:
         # NOTE: the following is unsafe
         # self.token_ids_cpu[i1, ...], self.token_ids_cpu[i2, ...], =\
         #     self.token_ids_cpu[i2, ...], self.token_ids_cpu[i1, ...]
-        # instead, we need to temporiarily copy the data for one of the indices
+        # instead, we need to temporarily copy the data for one of the indices
         # TODO(lucas): optimize this by only copying valid indices
         tmp = self.token_ids_cpu[i1, ...].copy()
         self.token_ids_cpu[i1, ...] = self.token_ids_cpu[i2, ...]