VibeVoice models

4 VibeVoice models from Microsoft on Hugging Face, from 323M to 8.7B parameters, published by Microsoft in F32 and BF16, with contexts from 64K to 128K tokens. The smallest official model, VibeVoice-ASR-BitNet, needs about 1.4 GB of VRAM at its published precision; the cheapest live fit is RTX 3070 at $0.088/hr.

Pick a size

One row per official VibeVoice size: the VRAM it needs at each precision, the cheapest GPU that holds it today, and the KV cache for a 32K-token context.

ModelParametersNative VRAMFP8 VRAMINT4 VRAMLive GPU fit (native)Est. $/hrKV cache at 32K
VibeVoice-ASR-BitNet323M1.4 GB0.4 GB0.2 GBRTX 3070$0.088/hrnot published for this architecture
VibeVoice-1.5B2.7B6.0 GB3.0 GB1.5 GBRTX 3070$0.088/hr0.88 GB
VibeVoice-ASR-HF8.3B18.6 GB9.3 GB4.7 GBRTX A5000$0.176/hr1.8 GB
VibeVoice-ASR8.7B19.4 GB9.7 GB4.8 GBRTX A5000$0.176/hrnot published for this architecture

VRAM is weights times a flat 1.2 overhead; the KV cache is a separate per-model figure at 16-bit, shown where the architecture is published. See the methodology.

Official models (4)

ModelParametersPublished asVRAM neededLive GPU fitEst. $/hr
VibeVoice-ASR-BitNet323MF321.4 GBRTX 3070$0.088/hr
VibeVoice-1.5B2.7BBF166.0 GBRTX 3070$0.088/hr
VibeVoice-ASR-HF8.3BBF1618.6 GBRTX A5000$0.176/hr
VibeVoice-ASR8.7BBF1619.4 GBRTX A5000$0.176/hr

VRAM is for the precision the model is published in: the weights times a flat 1.2 overhead, with the KV cache not included. See the methodology. The fit shown is the lowest-priced single GPU type that holds the model at that precision, or the lowest-priced multi-GPU set (up to 8) when none does.

More from Microsoft

Start ComfyUI on a rented GPU.

© 2026 Aquanode. All rights reserved.

All trademarks, logos and brand names are the property of their respective owners.