A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 9, 00:00 UTC
[Voxtral Realtime] Change name (#33716)
Signed-off-by: Patrick von Platen <patrick.v.platen@gmail.com>
Patrick von Platen committed
Feb 3, 2026 at 22:03 UTC
3f7662d6505e441026e668ba78a2207d669f4f32
4 files changed
+6
-6
examples/online_serving/openai_realtime_client.py
+2
-2
@@ -7,7 +7,7 @@ audio transcription by uploading an audio file.
7
Before running this script, you must start the vLLM server with a realtime-capable
8
model, for example:
9
10
- vllm serve mistralai/Voxtral-Mini-3B-Realtime-2602 --enforce-eager
10
+ vllm serve mistralai/Voxtral-Mini-4B-Realtime-2602 --enforce-eager
11
12
Requirements:
13
- vllm with audio support
@@ -126,7 +126,7 @@ if __name__ == "__main__":
126
parser.add_argument(
127
"--model",
128
type=str,
129
- default="mistralai/Voxtral-Mini-3B-Realtime-2602",
129
+ default="mistralai/Voxtral-Mini-4B-Realtime-2602",
130
help="Model that is served and should be pinged.",
131
)
132
parser.add_argument(
examples/online_serving/openai_realtime_microphone_client.py
+2
-2
@@ -5,7 +5,7 @@ Minimal Gradio demo for real-time speech transcription using the vLLM Realtime A
5
6
Start the vLLM server first:
7
8
- vllm serve mistralai/Voxtral-Mini-3B-Realtime-2602 --enforce-eager
8
+ vllm serve mistralai/Voxtral-Mini-4B-Realtime-2602 --enforce-eager
9
10
Then run this script:
11
@@ -166,7 +166,7 @@ if __name__ == "__main__":
166
parser.add_argument(
167
"--model",
168
type=str,
169
- default="mistralai/Voxtral-Mini-3B-Realtime-2602",
169
+ default="mistralai/Voxtral-Mini-4B-Realtime-2602",
170
help="Model that is served and should be pinged.",
171
)
172
parser.add_argument(
tests/entrypoints/openai/test_realtime_validation.py
+1
-1
@@ -24,7 +24,7 @@ MISTRAL_FORMAT_ARGS = [
24
"mistral",
25
]
26
27
-MODEL_NAME = "mistralai/Voxtral-Mini-3B-Realtime-2602"
27
+MODEL_NAME = "mistralai/Voxtral-Mini-4B-Realtime-2602"
28
29
30
def _audio_to_base64_pcm16(path: str, target_sr: int = 16000) -> str:
tests/models/multimodal/generation/test_voxtral_realtime.py
+1
-1
@@ -19,7 +19,7 @@ from vllm.engine.arg_utils import AsyncEngineArgs
19
from vllm.inputs.data import TokensPrompt
20
from vllm.v1.engine.async_llm import AsyncLLM, StreamingInput
21
22
-MODEL_NAME = "mistralai/Voxtral-Mini-3B-Realtime-2602"
22
+MODEL_NAME = "mistralai/Voxtral-Mini-4B-Realtime-2602"
23
ENGINE_CONFIG = dict(
24
model=MODEL_NAME,
25
max_model_len=8192,