Open-weight model · Text to speech
Ming-3.0-audio-tiny
by inclusionAI inclusionAI/Ming-3.0-audio-tiny
Ming-3.0-audio-tiny is an open-weight model for text to speech from inclusionAI, released under MIT License. It has 8.9B parameters. At 16-bit it needs about 21.3 GB of GPU memory, which fits on 1x MI300X from $1.85 an hour, at the lowest prices in the SAVRN Index. It draws 3 downloads a month.
Parameters8.9B
Context—
Weights—
Licensemit
AccessOpen weights
Monthly Downloads3
Runs On
What it takes to serve Ming-3.0-audio-tiny (8.9B parameters): the memory its weights need at each precision, and the cheapest way to rent enough data-center GPUs to hold them.
| Precision | Weights | Memory needed | Cheapest setup | Per hour | Also fits |
|---|---|---|---|---|---|
| 16-bit | 17.7 GB | 21.3 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 8-bit | 8.9 GB | 10.6 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 4-bit | 4.4 GB | 5.3 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
Memory is the weights at that precision plus 20% for the runtime and a short context; a long context needs more. Prices are the lowest on-demand hourly rates in the SAVRN Index, read Oct 20, 2026.
Ming-3.0-audio-tiny on every accelerator the SAVRN Index prices, at every precision
Model Card
Identity and Version
- Repository
- inclusionAI/Ming-3.0-audio-tiny
- Publisher
- inclusionAI
- Task
- Text to speech
- Modality
- Audio
- Library
- transformers
- Parameters
- 8.9B parameters
- Languages
- en, zh
- Revision
- 25b96e30be3739849bf76c0f48ac11db4a262fb3
- First published
- 2026-09-03
- Last updated
- 2026-10-20
License and Download
- License
- mit
- Access
- Open weights, no gate
Download from inclusionAI
Released by inclusionAI through its official repository on Hugging Face. Read the license.
Memory Requirements
| Precision | Weights in memory |
|---|---|
| 16-bit | 17.7 GB |
| 8-bit | 8.9 GB |
| 4-bit | 4.4 GB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.
Questions About Ming-3.0-audio-tiny
How much GPU memory does Ming-3.0-audio-tiny need?
About 21.3 GB at 16-bit and 5.3 GB at 4-bit: the weights (8.9B parameters) plus a working margin. A long context needs more.
What is the cheapest GPU to run Ming-3.0-audio-tiny on?
At 16-bit, 1x MI300X from $1.85 an hour; at 4-bit, 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists.
Can I use Ming-3.0-audio-tiny commercially?
Yes. Ming-3.0-audio-tiny is released under MIT License. The MIT License is a short permissive license. It permits commercial use, modification and redistribution, provided the copyright notice and permission notice are included.
Similar Models
MOSS-TTS-v1.5 is continued from MOSS-TTS 1.0. It preserves the main 1.0 capabilities, including zero-shot voice cloning, long-form speech generation, token-level duration control, Pinyin/IPA pronunciation control, multilingual synthesis, and code-switching. For the full 1.0 feature walkthrough, input schema, decoding hyperparameters, and evaluation tables, please refer to the MOSS-TTS 1.0 README. Compared with MOSS-TTS 1.0, v1.5 focuses on the following improvements: - Stronger multilingual synthesis with language tags: when the language field is omitted, v1.5 may improve some languages and regress slightly on others compared with 1.0. When the language is specified, v1.5 is stronger than…
The smallest working Drama 3. Unofficial int4 quantization of fishaudio/drama-3 — Fish Audio's 35B (3B active) multi-speaker TTS model — with real eval numbers, one-command serving, and a loader patch for SGLang-Omni. - 22.1 GB download (vs 41.5 GB original) — experts, attention and deltanet layers int4; shared-expert MLP kept bf16 (see What we quantize) - Runs on Ampere (A100, sm80) — the FP8 original needs Hopper/Ada (fp8e4nv kernels); our int4 kernels (moewna16 + GPTQ-Marlin) don't - First drama-3 quant with published WER + speaker-similarity eval (below) - 5 verified sample types: plain TTS, voice cloning, [sighs]/[laughs] tags, long-form, Chinese Compared against the strongest…
Check our open-source repository https://github.com/boson-ai/higgs-audio for more details! We are open-sourcing Higgs TTS 2, a powerful audio foundation model pretrained on over 10 million hours of audio data and a diverse set of text data. Despite having no post-training or fine-tuning, Higgs TTS 2 excels in expressive audio generation, thanks to its deep language and acoustic understanding. On EmergentTTS-Eval, the model achieves win rates of 75.7% and 55.7% over "gpt-4o-mini-tts" on the "Emotions" and "Questions" categories, respectively. It also obtains state-of-the-art performance on traditional TTS benchmarks like Seed-TTS Eval and Emotional Speech Dataset (ESD). Moreover, the model…
Higgs TTS 3 is built for voice chat: it speaks, not just reads. It turns model responses into expressive conversational speech across 100+ languages, with zero-shot voice cloning and inline control over emotion, style, prosody, pauses, and sound effects. Higgs autoregressive decoder consumes interleaved text and audio tokens. Audio is encoded by the Higgs Tokenizer into 8 codebooks at 25 fps, staggered via a delay pattern, then mapped to backbone hidden states through a multi-codebook fused embedding. Output codes pass through a multi-codebook fused head, are de-delayed, and decoded back to waveform. The model reaches single-digit WER/CER on 102 languages, which split into · Chichewa/Nyanja…
A trustremotecode packaging of bosonai/higgs-audio-v3-tts-4b that loads with plain transformers (no SGLang). The weights are the original checkpoint, copied unchanged; only a small modeling.py / configuration.py pair and an automap were added. The model is a standard Qwen3-4B backbone plus a fused multi-codebook audio embedding/head. Reference-audio encoding and waveform decoding use the transformers-native bosonai/higgs-audio-v2-tokenizer (higgsaudiov2tokenizer), loaded automatically on first use. Requires transformers >= 5.5. generatespeech returns a mono 24 kHz waveform as a CPU float32 tensor [L]. - Generation uses Higgs' delay pattern across 8 codebooks (vocab 1026, incl. BOC/EOC…