- Services
Ming-3.0-audio-flash is a native audio-language model from inclusionAI, the open source group behind Ant Group's Ming family. It takes audio plus text instructions and handles speech recognition, spoken question answering, audio understanding and audio reasoning, with optional synthesized speech output. It is a mixture-of-experts model with 126B total parameters and 6.5B activated per inference step, released under the MIT license. Vendor-reported scores include 82.0 on MMAU, 86.74 on VoiceBench, 99.4 on BigBenchAudio (S2T) and 82.6 on SpeechFCEval.
A situational 126B-parameter MoE audio model from inclusionAI. Treat the modality benchmarks above as the leading indicator of fit — composite scoring across modalities is still maturing. Newly released, so production-readiness is still being shaken out.
Generated from this model’s benchmarks and ranking signals. Editor reviews refine it over time.
Access model weights, configuration files, and documentation.
No benchmark data available for this model yet.
The top devices for this model at 4-bit, ranked by fit and speed.
| Device | Grade | VRAM |
|---|---|---|
| AMD Instinct MI300XAMD | SS | 80.2 GB |
| AMD Instinct MI325XAMD | SS | 80.2 GB |
| AMD Instinct MI355XAMD | SS | 80.2 GB |
| Apple M3 Ultra (32-core CPU, 80-core GPU)Apple | SS | 80.2 GB |
| Apple Mac Studio (M2 Ultra, 2023)Apple | SS | 80.2 GB |
Cheapest current cloud rentals with at least 80 GB VRAM, refreshed hourly.
Advertising disclosure: we earn commissions when you shop through the links below.
| Option | Cost / GPU-hour |
|---|---|
NVIDIA H100 NVLVast.ai · Spot · 94 GB VRAM | $0.99 |
NVIDIA H200 NVLVast.ai · Spot · 141 GB VRAM | $1.33 |
NVIDIA RTX PRO 6000 BlackwellRunPod · Community · 96 GB VRAM | $1.69 |
AMD Instinct MI300XRunPod · Secure · 192 GB VRAM | $2.39 |
NVIDIA RTX PRO 6000 BlackwellRunPod · Secure · 96 GB VRAM | $2.49 |
Per-GPU rate across RunPod, the Vast.ai marketplace, DigitalOcean, and Vultr.
Spot tier is interruptible. Plan for restarts when comparing against on-demand prices.
Ming-3.0-audio-flash combines an audio encoder and projector with a Mixture-of-Experts language backbone, plus an optional Talker that generates speech responses from the model output. It supports Chinese and English speech including dialects, accents, noisy environments and multi-speaker recordings, and can analyze long-form recordings covering speech, non-speech sounds, acoustic scenes and music. The repository ships custom Transformers code, a voice pack with two reference clips, and examples for audio understanding and ASR.
Explore the Provider
Aggregate stats, leaderboard, release timeline, and benchmark coverage across every inclusionAI model we track.
Check which of your devices can run this model and how fast.
Compare hosted per-token prices before you commit to local hardware.
Break-even math for self-hosting, renting a GPU, or paying per token.
Live status for the hosted APIs you might use instead of running this locally.
View on Hugging Face