Ming-3.0-audio-flash vs Underdog Bark 3.0

Ming-3.0-audio-flash

Ming-3.0-audio-flash is a cousin for Chinese and English transcription, questions about long audio and spoken replies in built-in or cloned voices, under MIT. It needs about 260 GB on a multi-GPU server, with inclusionAI's own runtime branches or Transformers remote code. The main risks are running repository code to load it and no published safety results.

Underdog Bark 3.0

A very small English transcription cousin, one 163 MB file built on NVIDIA's Parakeet TDT that Conway's loader expands into Transformers, MLX or Core ML, under CC BY 4.0. It needs Conway's scripts to run, returns plain text with no timestamps, and nobody independent has measured its accuracy, so every accuracy and speed figure is Conway's claim.

Ming-3.0-audio-flash

1. Start it with inclusionAI's recipe, hf download inclusionAI/Ming-3.0-audio-flash --local-dir inclusionAI/Ming-3.0-audio-flash && export MODEL_PATH="$PWD/inclusionAI/Ming-3.0-audio-flash" && pip install -r requirements-vllm-omni.txt && MING_OUTPUT_MODE=asr ./scripts/serve_vllm_omni.sh "$MODEL_PATH", run from a clone of github.com/inclusionAI/Ming-3.0-audio-flash
2. The 128K profile uses tensor parallelism over eight 80 GB H800 GPUs, the SGLang profile four and SGLang-Omni's cookbook two. It doesn't fit the 107.4 GB reference machine
3. Use MING_OUTPUT_MODE=asr with the validated prompt and greedy decoding for transcripts, and split the language label from the text at the tab
4. Use MING_OUTPUT_MODE=text for audio questions and the Talker profile with modalities=["text", "audio"] for spoken replies
5. Send audio as a base64 data URL in an audio_url item for vLLM-Omni, or as input_audio for SGLang
6. Pick a voice with audio.voice and an id, or clone one with ref_audio and its exact prompt_text, never both
7. Turn thinking on per request with chat_template_kwargs {"enable_thinking": true}, it's off by default and the reasoning comes back in think tags
8. Tool calls follow the template's <tool_call> with <arg_key> and <arg_value> pairs, and no parser is named, so parse them yourself
9. Review the repository's Python and the talker/ code, .pyc files included, before loading with trust_remote_code

Underdog Bark 3.0

1. Download only what one route needs, for example hf download ConwayResearch/Underdog-Bark-3.0 --include "model.bark" --include "bark_loader.py" --include "bark_mlx.py" --include "mlx/*" --include "NOTICE"
2. With transformers 5.19.0 and constriction 0.5.0, python bark_loader.py model.bark recording.wav, or load_model("model.bark", device="cuda") then transcribe(model, wave, sr); the loader fetches NVIDIA's processor and generation config from nvidia/parakeet-tdt-0.6b-v3 on first use, so the first run needs the network
3. On Apple silicon, python bark_mlx.py recording.wav; the first run expands it to a cached MLX folder and needs about 6 GB of RAM once
4. Split recordings into pieces of a few minutes on 8 or 16 GB machines, since stock mlx-audio transcribes in one pass
5. Read bark_loader.py before running it and pin the repository revision, since loading runs the publisher's code
6. Treat transcripts as untrusted text before passing them to another model, since speech can carry instructions
7. Plan memory for the recording, not the file. Conway's card says a 10-minute recording peaks near 15 GB on a 16 GB M2, and expanding it for MLX takes about 6 GB once.