diff --git a/src/rendering/twn_draw_c.h b/src/rendering/twn_draw_c.h
index 9125d7b..d618c01 100644
--- a/src/rendering/twn_draw_c.h
+++ b/src/rendering/twn_draw_c.h
@@ -341,5 +341,7 @@ void issue_deferred_draw_commands(void);
 
 bool model_load_workers_thread(void);
 void finally_draw_models(void);
+void free_model_cache(void);
+void model_state_deinit(void);
 
 #endif
diff --git a/src/rendering/twn_model.c b/src/rendering/twn_model.c
index 101f054..4d3e61b 100644
--- a/src/rendering/twn_model.c
+++ b/src/rendering/twn_model.c
@@ -1,6 +1,5 @@
 #include "twn_draw_c.h"
 #include "twn_draw.h"
-#include "twn_util.h"
 #include "twn_workers_c.h"
 
 #define FAST_OBJ_IMPLEMENTATION
@@ -13,7 +12,7 @@
 
 
 static struct ModelCacheItem {
-    char const *key;
+    char *key;
     struct ModelCacheItemValue {
         fastObjMesh *mesh;
     } value;
@@ -29,7 +28,7 @@ static struct ModelDrawCommand {
 
 /* deferred queue of model files to load from worker threads */
 static SDL_mutex *model_load_mutex;
-static char **model_load_queue;
+static char const **model_load_queue;
 static bool model_load_initialized;
 
 /* use streaming via callbacks to reduce memory congestion */
@@ -69,7 +68,7 @@ static bool model_load_workers_finished(void) {
 /* returns false if there was nothing to do */
 bool model_load_workers_thread(void) {
     /* attempt to grab something to work on */
-    char *load_request = NULL;
+    char const *load_request = NULL;
     SDL_LockMutex(model_load_mutex);
     if (arrlenu(model_load_queue) != 0)
         load_request = arrpop(model_load_queue);
@@ -107,26 +106,26 @@ void draw_model(const char *model,
         model_load_initialized = true;
     }
 
-    /* make sure not to reference parameter longer than duration of this function */
-    char *model_copy = SDL_strdup(model);
-    model = NULL; /* trap */
+    struct ModelCacheItem const *item;
+
+    /* if model is missing, queue it up for loading */
+    SDL_LockMutex(model_load_mutex);
+    if (!(item = shgetp_null(model_cache, model))) {
+        model = SDL_strdup(model);
+        shput(model_cache, model, (struct ModelCacheItemValue){0});
+        arrpush(model_load_queue, model);
+        SDL_SemPost(workers_job_semaphore);
+    } else
+        model = item->key;
+    SDL_UnlockMutex(model_load_mutex);
 
     struct ModelDrawCommand const command = {
-        .model = model_copy,
+        .model = (char *)model,
         .position = position,
         .rotation = rotation,
         .scale = scale
     };
     arrpush(model_draw_commands, command);
-
-    /* if model is missing, queue it up for loading */
-    SDL_LockMutex(model_load_mutex);
-    if (!(shgetp_null(model_cache, model_copy))) {
-        shput(model_cache, model_copy, (struct ModelCacheItemValue){0});
-        arrpush(model_load_queue, model_copy);
-        SDL_SemPost(workers_job_semaphore);
-    }
-    SDL_UnlockMutex(model_load_mutex);
 }
 
 
@@ -147,20 +146,57 @@ void finally_draw_models(void) {
                 // fastObjTexture const *const texture = &mesh->textures[group->face_offset + f];
                 // log_info("material: %s", material->name);
                 /* TODO: support arbitrary fans */
-                SDL_assert(vertices == 4);
-                fastObjIndex i0 = mesh->indices[group->index_offset + idx + 0];
-                fastObjIndex i1 = mesh->indices[group->index_offset + idx + 1];
-                fastObjIndex i2 = mesh->indices[group->index_offset + idx + 2];
-                fastObjIndex i3 = mesh->indices[group->index_offset + idx + 3];
-                draw_quad(
-                    "asd",
-                    (Vec3) { mesh->positions[3 * i0.p + 0], mesh->positions[3 * i0.p + 1], mesh->positions[3 * i0.p + 2] },
-                    (Vec3) { mesh->positions[3 * i1.p + 0], mesh->positions[3 * i1.p + 1], mesh->positions[3 * i1.p + 2] },
-                    (Vec3) { mesh->positions[3 * i2.p + 0], mesh->positions[3 * i2.p + 1], mesh->positions[3 * i2.p + 2] },
-                    (Vec3) { mesh->positions[3 * i3.p + 0], mesh->positions[3 * i3.p + 1], mesh->positions[3 * i3.p + 2] },
-                    (Rect) { .w = 64, .h = 64 },
-                    (Color) { 255, 255, 255, 255 }
-                );
+                unsigned int const material_index = mesh->face_materials[group->index_offset + f];
+                fastObjMaterial const *const material = mesh->materials ? &mesh->materials[material_index] : NULL;
+                if (vertices == 4) {
+                    fastObjIndex const i0 = mesh->indices[group->index_offset + idx + 0];
+                    fastObjIndex const i1 = mesh->indices[group->index_offset + idx + 1];
+                    fastObjIndex const i2 = mesh->indices[group->index_offset + idx + 2];
+                    fastObjIndex const i3 = mesh->indices[group->index_offset + idx + 3];
+                    draw_quad(
+                        material ? mesh->textures[material->map_Kd].name : NULL,
+                        (Vec3) { mesh->positions[3 * i0.p + 0] * command->scale.x, mesh->positions[3 * i0.p + 1] * command->scale.y, mesh->positions[3 * i0.p + 2] * command->scale.z },
+                        (Vec3) { mesh->positions[3 * i1.p + 0] * command->scale.x, mesh->positions[3 * i1.p + 1] * command->scale.y, mesh->positions[3 * i1.p + 2] * command->scale.z },
+                        (Vec3) { mesh->positions[3 * i2.p + 0] * command->scale.x, mesh->positions[3 * i2.p + 1] * command->scale.y, mesh->positions[3 * i2.p + 2] * command->scale.z },
+                        (Vec3) { mesh->positions[3 * i3.p + 0] * command->scale.x, mesh->positions[3 * i3.p + 1] * command->scale.y, mesh->positions[3 * i3.p + 2] * command->scale.z },
+                        (Rect) { .w = 64, .h = 64 },
+                        (Color) { 255, 255, 255, 255 }
+                    );
+                } else if (vertices == 3) {
+                    fastObjIndex const i0 = mesh->indices[group->index_offset + idx + 0];
+                    fastObjIndex const i1 = mesh->indices[group->index_offset + idx + 1];
+                    fastObjIndex const i2 = mesh->indices[group->index_offset + idx + 2];
+                    draw_triangle(
+                        material ? mesh->textures[material->map_Kd].name : NULL,
+                        (Vec3) { mesh->positions[3 * i0.p + 0] * command->scale.x, mesh->positions[3 * i0.p + 1] * command->scale.y, mesh->positions[3 * i0.p + 2] * command->scale.z },
+                        (Vec3) { mesh->positions[3 * i1.p + 0] * command->scale.x, mesh->positions[3 * i1.p + 1] * command->scale.y, mesh->positions[3 * i1.p + 2] * command->scale.z },
+                        (Vec3) { mesh->positions[3 * i2.p + 0] * command->scale.x, mesh->positions[3 * i2.p + 1] * command->scale.y, mesh->positions[3 * i2.p + 2] * command->scale.z },
+                        (Vec2) {0,0},
+                        (Vec2) {0,0},
+                        (Vec2) {0,0},
+                        (Color){255, 255, 255, 255},
+                        (Color){255, 255, 255, 255},
+                        (Color){255, 255, 255, 255}
+                    );
+                } else {
+                    fastObjIndex const i0 = mesh->indices[group->index_offset + idx];
+                    for (unsigned int z = 0; z < vertices - 2; ++z) {
+                        fastObjIndex const i1 = mesh->indices[group->index_offset + idx + 1 + z];
+                        fastObjIndex const i2 = mesh->indices[group->index_offset + idx + 2 + z];
+                        draw_triangle(
+                            material ? mesh->textures[material->map_Kd].name : NULL,
+                            (Vec3) { mesh->positions[3 * i0.p + 0] * command->scale.x, mesh->positions[3 * i0.p + 1] * command->scale.y, mesh->positions[3 * i0.p + 2] * command->scale.z },
+                            (Vec3) { mesh->positions[3 * i1.p + 0] * command->scale.x, mesh->positions[3 * i1.p + 1] * command->scale.y, mesh->positions[3 * i1.p + 2] * command->scale.z },
+                            (Vec3) { mesh->positions[3 * i2.p + 0] * command->scale.x, mesh->positions[3 * i2.p + 1] * command->scale.y, mesh->positions[3 * i2.p + 2] * command->scale.z },
+                            (Vec2) {0,0},
+                            (Vec2) {0,0},
+                            (Vec2) {0,0},
+                            (Color){255, 255, 255, 255},
+                            (Color){255, 255, 255, 255},
+                            (Color){255, 255, 255, 255}
+                        );
+                    }
+                }
                 idx += vertices;
             }
         }
@@ -168,3 +204,27 @@ void finally_draw_models(void) {
 
     arrsetlen(model_draw_commands, 0);
 }
+
+
+/* drop model caches */
+void free_model_cache(void) {
+    while (!model_load_workers_finished()) {
+        (void)0;
+    }
+
+    for (size_t i = 0; i < shlenu(model_cache); ++i) {
+        fast_obj_destroy(model_cache[i].value.mesh);
+        SDL_free(model_cache[i].key);
+    }
+
+    shfree(model_cache);
+}
+
+
+void model_state_deinit(void) {
+    SDL_assert(model_load_initialized);
+    free_model_cache();
+    arrfree(model_load_queue);
+    SDL_DestroyMutex(model_load_mutex);
+    model_load_initialized = false;
+}
diff --git a/src/twn_loop.c b/src/twn_loop.c
index aa1b73a..6c4378e 100644
--- a/src/twn_loop.c
+++ b/src/twn_loop.c
@@ -726,6 +726,7 @@ static void clean_up(void) {
     toml_free(ctx.config_table);
     PHYSFS_deinit();
     workers_deinit();
+    model_state_deinit();
     SDL_free(ctx.base_dir);
     SDL_free(ctx.title);
     SDL_GL_DeleteContext(ctx.gl_context);
diff --git a/src/twn_workers.c b/src/twn_workers.c
index 7dd64cd..e4a1a18 100644
--- a/src/twn_workers.c
+++ b/src/twn_workers.c
@@ -4,10 +4,10 @@
 
 SDL_sem *workers_job_semaphore;
 
-static SDL_Thread *workers_pool[MAX_WORKERS];
 static size_t workers_pool_size;
 static SDL_mutex *workers_mutex;
 static bool workers_should_exit;
+static SDL_sem *workers_exit_semaphore; /* should come to count of `workers_pool_size` */
 
 /* logic is such that when job is posted, worker threads attempt to grab it from any possible entry point */
 /* if it did something, which is signaled by `true` return, go back to waiting on semaphore, so that it's decremented properly */
@@ -31,6 +31,9 @@ static int worker_thread(void *udata) {
             continue;
     }
 
+    /* let the main thread collect it */
+    SDL_SemPost(workers_exit_semaphore);
+
     return 0;
 }
 
@@ -52,6 +55,7 @@ bool workers_init(size_t worker_count) {
 
     workers_pool_size = worker_count;
     workers_job_semaphore = SDL_CreateSemaphore(0);
+    workers_exit_semaphore = SDL_CreateSemaphore(0);
     workers_mutex = SDL_CreateMutex();
     return true;
 }
@@ -62,8 +66,12 @@ void workers_deinit(void) {
     workers_should_exit = true;
     SDL_UnlockMutex(workers_mutex);
 
-    /* TODO: that's not correct */
+    for (size_t i = 0; i < workers_pool_size; ++i) {
+        SDL_SemWait(workers_exit_semaphore);
+    }
+
     SDL_DestroyMutex(workers_mutex);
     SDL_DestroySemaphore(workers_job_semaphore);
+    SDL_DestroySemaphore(workers_exit_semaphore);
     workers_pool_size = 0;
 }