diff --git a/src/pse/gpu.cpp b/src/pse/gpu.cpp
index 23d28eee7..7a2edc668 100644
--- a/src/pse/gpu.cpp
+++ b/src/pse/gpu.cpp
@@ -322,7 +322,7 @@ bool GPU::HandleRenderCommand()
     case Primitive::Rectangle:
     {
       words_per_vertex =
-        1 + BoolToUInt8(rc.texture_enable) + BoolToUInt8(rc.rectangle_size == DrawRectangleSize::Variable);
+        2 + BoolToUInt8(rc.texture_enable) + BoolToUInt8(rc.rectangle_size == DrawRectangleSize::Variable);
       num_vertices = 1;
       total_words = words_per_vertex;
     }
diff --git a/src/pse/gpu_hw.cpp b/src/pse/gpu_hw.cpp
index eaf602d37..f0b941f00 100644
--- a/src/pse/gpu_hw.cpp
+++ b/src/pse/gpu_hw.cpp
@@ -8,6 +8,7 @@ GPU_HW::~GPU_HW() = default;
 
 void GPU_HW::LoadVertices(RenderCommand rc, u32 num_vertices)
 {
+  // TODO: Move this to the GPU..
   switch (rc.primitive)
   {
     case Primitive::Polygon:
@@ -48,6 +49,53 @@ void GPU_HW::LoadVertices(RenderCommand rc, u32 num_vertices)
     }
     break;
 
+    case Primitive::Rectangle:
+    {
+      u32 buffer_pos = 1;
+
+      const bool textured = rc.texture_enable;
+      const u32 color = rc.color_for_first_vertex;
+      const VertexPosition vp{m_GP0_command[buffer_pos++]};
+      const s32 pos_left = vp.x();
+      const s32 pos_top = vp.y();
+      const auto [tex_left, tex_top] =
+        HWVertex::DecodeTexcoord(rc.texture_enable ? Truncate16(m_GP0_command[buffer_pos++]) : 0);
+      s32 rectangle_width;
+      s32 rectangle_height;
+      switch (rc.rectangle_size)
+      {
+        case DrawRectangleSize::R1x1:
+          rectangle_width = 1;
+          rectangle_height = 1;
+          break;
+        case DrawRectangleSize::R8x8:
+          rectangle_width = 8;
+          rectangle_height = 8;
+          break;
+        case DrawRectangleSize::R16x16:
+          rectangle_width = 16;
+          rectangle_height = 16;
+          break;
+        default:
+          rectangle_width = static_cast<s32>(m_GP0_command[buffer_pos] & UINT32_C(0xFFFF));
+          rectangle_height = static_cast<s32>(m_GP0_command[buffer_pos] >> 16);
+          break;
+      }
+
+      // TODO: This should repeat the texcoords instead of stretching
+      const s32 pos_right = pos_left + (rectangle_width - 1);
+      const s32 pos_bottom = pos_top + (rectangle_height - 1);
+      const u8 tex_right = static_cast<u8>(tex_left + (rectangle_width - 1));
+      const u8 tex_bottom = static_cast<u8>(tex_top + (rectangle_height - 1));
+
+      m_batch_vertices.push_back(HWVertex{pos_left, pos_top, color, HWVertex::EncodeTexcoord(tex_left, tex_top)});
+      m_batch_vertices.push_back(HWVertex{pos_right, pos_top, color, HWVertex::EncodeTexcoord(tex_right, tex_top)});
+      m_batch_vertices.push_back(HWVertex{pos_left, pos_bottom, color, HWVertex::EncodeTexcoord(tex_left, tex_bottom)});
+      m_batch_vertices.push_back(
+        HWVertex{pos_right, pos_bottom, color, HWVertex::EncodeTexcoord(tex_right, tex_bottom)});
+    }
+    break;
+
     default:
       UnreachableCode();
       break;
@@ -272,6 +320,12 @@ void GPU_HW::DispatchRenderCommand(RenderCommand rc, u32 num_vertices)
         else
           m_texture_config.SetFromPolygonTexcoord(m_GP0_command[2], m_GP0_command[4]);
       }
+
+      case Primitive::Rectangle:
+      {
+        m_texture_config.SetFromRectangleTexcoord(m_GP0_command[2]);
+        m_texture_config.SetFromPageAttribute(Truncate16(m_GPUSTAT.bits));
+      }
       break;
 
       default:
diff --git a/src/pse/gpu_hw.h b/src/pse/gpu_hw.h
index 3a8f96f5b..aa054f4bb 100644
--- a/src/pse/gpu_hw.h
+++ b/src/pse/gpu_hw.h
@@ -1,7 +1,7 @@
 #pragma once
 #include "gpu.h"
-#include <string>
 #include <sstream>
+#include <string>
 #include <vector>
 
 class GPU_HW : public GPU
@@ -18,6 +18,12 @@ protected:
     u32 color;
     u16 texcoord;
     u16 padding;
+
+    static std::tuple<u8, u8> DecodeTexcoord(u16 texcoord)
+    {
+      return std::make_tuple(static_cast<u8>(texcoord), static_cast<u8>(texcoord >> 8));
+    }
+    static u16 EncodeTexcoord(u8 x, u8 y) { return ZeroExtend16(x) | (ZeroExtend16(y) << 8); }
   };
 
   virtual void UpdateTexturePageTexture();
@@ -42,4 +48,3 @@ private:
 
   void LoadVertices(RenderCommand rc, u32 num_vertices);
 };
-
diff --git a/src/pse/gpu_hw_opengl.cpp b/src/pse/gpu_hw_opengl.cpp
index 2aa1d854b..256498564 100644
--- a/src/pse/gpu_hw_opengl.cpp
+++ b/src/pse/gpu_hw_opengl.cpp
@@ -283,8 +283,9 @@ void GPU_HW_OpenGL::FlushRender()
   glVertexAttribPointer(2, 2, GL_UNSIGNED_BYTE, true, sizeof(HWVertex),
                         reinterpret_cast<void*>(offsetof(HWVertex, texcoord)));
 
-  glDrawArrays(m_batch_command.quad_polygon ? GL_TRIANGLE_STRIP : GL_TRIANGLES, 0,
-               static_cast<GLsizei>(m_batch_vertices.size()));
+  const bool is_strip = ((m_batch_command.primitive == Primitive::Polygon && m_batch_command.quad_polygon) ||
+                         m_batch_command.primitive == Primitive::Rectangle);
+  glDrawArrays(is_strip ? GL_TRIANGLE_STRIP : GL_TRIANGLES, 0, static_cast<GLsizei>(m_batch_vertices.size()));
 
   m_batch_vertices.clear();
 }