diff --git a/src/llama-model.cpp b/src/llama-model.cpp
index 5e37e03331..3881621df0 100644
--- a/src/llama-model.cpp
+++ b/src/llama-model.cpp
@@ -112,7 +112,7 @@ static llama_model * llama_model_mapping(llm_arch arch, const llama_model_params
         case LLM_ARCH_QWEN3VLMOE:
             return new llama_model_qwen3vlmoe(params);
         case LLM_ARCH_QWEN3TTS:
-            return new llama_model_qwen3vl(params);
+            return new llama_model_qwen3tts(params);
         case LLM_ARCH_PHI2:
             return new llama_model_phi2(params);
         case LLM_ARCH_PHI3:
diff --git a/src/models/models.h b/src/models/models.h
index bb372ece81..58556102f0 100644
--- a/src/models/models.h
+++ b/src/models/models.h
@@ -596,6 +596,11 @@ struct llama_model_qwen3vlmoe : public llama_model_base {
 };
 
 
+struct llama_model_qwen3tts : public llama_model_qwen3vl {
+    llama_model_qwen3tts(const struct llama_model_params & params) : llama_model_qwen3vl(params) {}
+};
+
+
 struct llama_model_phi2 : public llama_model_base {
     llama_model_phi2(const struct llama_model_params & params) : llama_model_base(params) {}
     void load_arch_hparams(llama_model_loader & ml) override;
diff --git a/src/models/qwen3tts.cpp b/src/models/qwen3tts.cpp
new file mode 100644
index 0000000000..3604f844c3
--- /dev/null
+++ b/src/models/qwen3tts.cpp
@@ -0,0 +1,3 @@
+#include "models.h"
+
+// llama_model_qwen3tts reuses llama_model_qwen3vl's hparams/tensors/graph logic
diff --git a/tests/test-llama-archs.cpp b/tests/test-llama-archs.cpp
index a1ed2a76f8..eb90798b62 100644
--- a/tests/test-llama-archs.cpp
+++ b/tests/test-llama-archs.cpp
@@ -113,6 +113,8 @@ static gguf_context_ptr get_gguf_ctx(const llm_arch arch, const bool moe) {
         n_layer = 3;
     } else if (arch == LLM_ARCH_CHAMELEON) {
         n_vocab = 10240;
+    } else if (arch == LLM_ARCH_QWEN3TTS) {
+        n_vocab = 4096; // must be >= the hard-coded codec head size (3072)
     }
 
     const uint32_t n_embd_head = n_embd / n_head;