diff --git a/docs/contributing/model/transcription.md b/docs/contributing/model/transcription.md
index fca941acd..7fe010e5f 100644
--- a/docs/contributing/model/transcription.md
+++ b/docs/contributing/model/transcription.md
@@ -251,6 +251,7 @@ No extra registration is required beyond having your model class available via t
 - Whisper encoder–decoder (audio-only): [vllm/model_executor/models/whisper.py](../../../vllm/model_executor/models/whisper.py)
 - Voxtral decoder-only (audio embeddings + LLM): [vllm/model_executor/models/voxtral.py](../../../vllm/model_executor/models/voxtral.py). Make sure to have installed `mistral-common[audio]`.
 - Gemma3n decoder-only with fixed instruction prompt: [vllm/model_executor/models/gemma3n_mm.py](../../../vllm/model_executor/models/gemma3n_mm.py)
+- Qwen3-Omni multimodal with audio embeddings: [vllm/model_executor/models/qwen3_omni_moe_thinker.py](../../../vllm/model_executor/models/qwen3_omni_moe_thinker.py)
 
 ## Test with the API
 
diff --git a/docs/models/supported_models.md b/docs/models/supported_models.md
index a96abd891..e07e17ec5 100644
--- a/docs/models/supported_models.md
+++ b/docs/models/supported_models.md
@@ -781,6 +781,7 @@ Speech2Text models trained specifically for Automatic Speech Recognition.
 | `GlmAsrForConditionalGeneration` | GLM-ASR | `zai-org/GLM-ASR-Nano-2512` | ✅︎ | ✅︎ |
 | `GraniteSpeechForConditionalGeneration` | Granite Speech | `ibm-granite/granite-speech-3.3-2b`, `ibm-granite/granite-speech-3.3-8b`, etc. | ✅︎ | ✅︎ |
 | `Qwen3ASRForConditionalGeneration` | Qwen3-ASR | `Qwen/Qwen3-ASR-1.7B`, etc. | | ✅︎ |
+| `Qwen3OmniMoeThinkerForConditionalGeneration` | Qwen3-Omni | `Qwen/Qwen3-Omni-30B-A3B-Instruct`, etc. | | ✅︎ |
 | `VoxtralForConditionalGeneration` | Voxtral (Mistral format) | `mistralai/Voxtral-Mini-3B-2507`, `mistralai/Voxtral-Small-24B-2507`, etc. | ✅︎ | ✅︎ |
 | `WhisperForConditionalGeneration` | Whisper | `openai/whisper-small`, `openai/whisper-large-v3-turbo`, etc. | | |
 
diff --git a/vllm/model_executor/models/qwen3_omni_moe_thinker.py b/vllm/model_executor/models/qwen3_omni_moe_thinker.py
index 93a17f0c8..b06503031 100755
--- a/vllm/model_executor/models/qwen3_omni_moe_thinker.py
+++ b/vllm/model_executor/models/qwen3_omni_moe_thinker.py
@@ -24,7 +24,7 @@
 
 from collections.abc import Callable, Iterable, Iterator, Mapping, Sequence
 from functools import partial
-from typing import Any
+from typing import Any, Literal, cast
 
 import numpy as np
 import torch
@@ -48,8 +48,9 @@ from transformers import __version__ as TRANSFORMERS_VERSION
 # isort: on
 
 from vllm.compilation.decorators import support_torch_compile
-from vllm.config import VllmConfig
+from vllm.config import ModelConfig, SpeechToTextConfig, VllmConfig
 from vllm.distributed import get_pp_group, get_tensor_model_parallel_world_size
+from vllm.inputs.data import PromptType
 from vllm.logger import init_logger
 from vllm.model_executor.layers.activation import _ACTIVATION_REGISTRY
 from vllm.model_executor.layers.attention.mm_encoder_attention import (
@@ -79,6 +80,7 @@ from vllm.multimodal.processing.processor import (
     PromptUpdateDetails,
 )
 from vllm.sequence import IntermediateTensors
+from vllm.transformers_utils.processor import cached_processor_from_config
 from vllm.v1.attention.backends.registry import AttentionBackendEnum
 
 from .interfaces import (
@@ -86,6 +88,7 @@ from .interfaces import (
     SupportsMRoPE,
     SupportsMultiModal,
     SupportsPP,
+    SupportsTranscription,
 )
 from .qwen2_5_omni_thinker import (
     Qwen2_5OmniAudioFeatureInputs,
@@ -110,6 +113,29 @@ from .vision import get_vit_attn_backend
 
 logger = init_logger(__name__)
 
+# Speech input languages supported by Qwen3-Omni
+# From: https://huggingface.co/Qwen/Qwen3-Omni-30B-A3B-Instruct
+ISO639_1_SUPPORTED_LANGS = {
+    "en": "English",
+    "zh": "Chinese",
+    "ko": "Korean",
+    "ja": "Japanese",
+    "de": "German",
+    "ru": "Russian",
+    "it": "Italian",
+    "fr": "French",
+    "es": "Spanish",
+    "pt": "Portuguese",
+    "ms": "Malay",
+    "nl": "Dutch",
+    "id": "Indonesian",
+    "tr": "Turkish",
+    "vi": "Vietnamese",
+    "yue": "Cantonese",
+    "ar": "Arabic",
+    "ur": "Urdu",
+}
+
 
 def _get_feat_extract_output_lengths(input_lengths: torch.Tensor):
     input_lengths_leave = input_lengths % 100
@@ -1572,6 +1598,7 @@ class Qwen3OmniMoeThinkerForConditionalGeneration(
     SupportsPP,
     SupportsMRoPE,
     Qwen3OmniMoeConditionalGenerationMixin,
+    SupportsTranscription,
 ):
     hf_to_vllm_mapper = WeightsMapper(
         orig_to_new_prefix={
@@ -1593,6 +1620,8 @@ class Qwen3OmniMoeThinkerForConditionalGeneration(
         ],
     }
 
+    supported_languages = ISO639_1_SUPPORTED_LANGS
+
     @classmethod
     def get_placeholder_str(cls, modality: str, i: int) -> str | None:
         if modality.startswith("image"):
@@ -2085,6 +2114,77 @@ class Qwen3OmniMoeThinkerForConditionalGeneration(
         total_tokens = num_video + audio_len
         return np.concatenate(pos_ids_list, axis=1), total_tokens
 
+    @classmethod
+    def get_speech_to_text_config(
+        cls, model_config: ModelConfig, task_type: str
+    ) -> SpeechToTextConfig:
+        processor = cached_processor_from_config(
+            model_config, processor_cls=Qwen3OmniMoeProcessor
+        )
+        return SpeechToTextConfig(
+            max_audio_clip_s=processor.feature_extractor.chunk_length,
+            sample_rate=processor.feature_extractor.sampling_rate,
+            min_energy_split_window_size=None,
+        )
+
+    @classmethod
+    def get_generation_prompt(
+        cls,
+        audio: np.ndarray,
+        stt_config: SpeechToTextConfig,
+        model_config: ModelConfig,
+        language: str | None,
+        task_type: Literal["transcribe", "translate"],
+        request_prompt: str,
+        to_language: str | None,
+    ) -> PromptType:
+        """
+        Construct a transcription/translation prompt for Qwen3-Omni.
+        """
+        # Transcribe this audio [into <language>] | for transcription
+        # Translate this audio [from <language> into <to_language>] | for translation
+        instruction = "Transcribe" if task_type == "transcribe" else "Translate"
+        instruction += " this audio"
+
+        # Default to_language to English for translation
+        if task_type == "translate" and to_language is None:
+            to_language = "en"
+
+        # Get full language names from supported_languages mapping
+        full_lang_name = cls.supported_languages.get(language, "")
+        full_lang_name_to = cls.supported_languages.get(to_language, "")
+
+        if task_type == "transcribe" and full_lang_name:
+            instruction += f" into {full_lang_name}"
+        elif task_type == "translate":
+            if full_lang_name:
+                instruction += f" from {full_lang_name}"
+            if full_lang_name_to:
+                instruction += f" into {full_lang_name_to}"
+
+        instruction += "."
+
+        if request_prompt:
+            instruction += f" {request_prompt}"
+
+        processor = cached_processor_from_config(
+            model_config, processor_cls=Qwen3OmniMoeProcessor
+        )
+        # Audio placeholder format: <|audio_start|><|audio_pad|><|audio_end|>
+        audio_placeholder = "<|audio_start|><|audio_pad|><|audio_end|>"
+        user_content = f"{audio_placeholder}{instruction}"
+
+        messages = [{"role": "user", "content": user_content}]
+        prompt = processor.apply_chat_template(
+            messages,
+            tokenize=False,
+            add_generation_prompt=True,
+        )
+
+        audio_data = (audio, stt_config.sample_rate)
+        prompts_dict = {"multi_modal_data": {"audio": audio_data}, "prompt": prompt}
+        return cast(PromptType, prompts_dict)
+
     def get_mrope_input_positions(
         self,
         input_tokens: list[int],