# VibeVoice-ASR by Microsoft: Open-Weight Model Source: https://savrn.com/models/vibevoice-asr Markdown alternate of the page above; the site index is https://savrn.com/llms.txt --- ## Runs On What it takes to serve VibeVoice-ASR (8.7B 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 | 17.3 GB | 20.8 GB | 1x [MI300X](https://savrn.com/ai-index/pricing/gpus/mi300x) (192 GB) Vultr | $1.85 | [1x H100](https://savrn.com/ai-index/pricing/gpus/h100) $1.99 · [1x MI325X](https://savrn.com/ai-index/pricing/gpus/mi325x) $2.00 | | 8-bit | 8.7 GB | 10.4 GB | 1x [MI300X](https://savrn.com/ai-index/pricing/gpus/mi300x) (192 GB) Vultr | $1.85 | [1x H100](https://savrn.com/ai-index/pricing/gpus/h100) $1.99 · [1x MI325X](https://savrn.com/ai-index/pricing/gpus/mi325x) $2.00 | | 4-bit | 4.3 GB | 5.2 GB | 1x [MI300X](https://savrn.com/ai-index/pricing/gpus/mi300x) (192 GB) Vultr | $1.85 | [1x H100](https://savrn.com/ai-index/pricing/gpus/h100) $1.99 · [1x MI325X](https://savrn.com/ai-index/pricing/gpus/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](https://savrn.com/ai-index/pricing/gpus), read Oct 24, 2026. [VibeVoice-ASR on every accelerator the SAVRN Index prices, at every precision](https://savrn.com/models/vibevoice-asr/gpus) ## SAVRN's Notes on VibeVoice-ASR If you need who said what and when from a long recording, this is the model we would size first. VibeVoice-ASR takes up to 60 minutes of continuous audio in a single pass inside a 64K token length and returns speaker, timestamp and content together, with custom hotwords and more than 50 languages. The 8.7 billion parameters weigh 17.3 GB at 16-bit and need 20.8 GB of memory, which fits one MI300X with 192 GB at $1.85 an hour; 8-bit needs 10.4 GB and 4-bit 5.2 GB. The rest of the card is headroom for parallel streams. MIT is the license, dated 2026-01-21 from Microsoft, which clears commercial transcription products, changes and redistribution as long as the notice travels with the files. Before committing, note that the page lists no context length, so the publisher's 60-minute, 64K token figure is your ceiling, and that no SAVRN Index host prices it, so plan on your own card. SAVRN Research, 2026-09-18 ## Model Card By Microsoft, published under mit, revision d0c9efdb8d61. VibeVoice-ASR is a unified speech-to-text model designed to handle 60-minute long-form audio in a single pass, generating structured transcriptions containing Who (Speaker), When (Timestamps), and What (Content), with support for Customized Hotwords and over 50 languages. Code: [microsoft/VibeVoice](https://github.com/microsoft/VibeVoice) Demo: [VibeVoice-ASR-Demo](https://aka.ms/vibevoice-asr) Report: [VibeVoice-ASR Technical Report](https://arxiv.org/pdf/2601.18184) Finetuning: [Finetuning](https://github.com/microsoft/VibeVoice/blob/main/finetuning-asr/README.md) vLLM: [vLLM-VibeVoice-ASR](https://github.com/microsoft/VibeVoice/blob/main/docs/vibevoice-vllm-asr.md) ### Key Features [Read the full model card (255 words)](https://savrn.com/models/vibevoice-asr/card) ## Configuration Architecture VibeVoiceForASRTraining Model type vibevoice ## Identity and Version Repository microsoft/VibeVoice-ASR Publisher Microsoft Task Speech recognition Modality Audio Library transformers Parameters 8.7B parameters Languages en, zh, es, pt, de, ja, ko, fr Revision d0c9efdb8d614685062c04425d91e01b6f37d944 First published 2026-01-21 Last updated 2026-01-27 ## Files and Weights 17 files, 17.3 GB in total. The weights are 8 files totalling 17.3 GB in safetensors. Weights8 files · 17.3 GB Configuration2 files · 123.7 KB Documentation1 file · 4.3 KB Other5 files · 1.2 MB Repository1 file · 199 B Every file | File | Type | Size | SHA-256 | | --- | --- | --- | --- | | model-00001-of-00008.safetensors | Weights | 2.5 GB | 5548c67885d4 | | model-00002-of-00008.safetensors | Weights | 2.4 GB | 163023c61a3f | | model-00003-of-00008.safetensors | Weights | 2.5 GB | 4e021702dfac | | model-00004-of-00008.safetensors | Weights | 2.5 GB | b17657bb151d | | model-00005-of-00008.safetensors | Weights | 2.5 GB | 0ed4e457268f | | model-00006-of-00008.safetensors | Weights | 2.5 GB | 6de8246bb042 | | model-00007-of-00008.safetensors | Weights | 1.5 GB | a2ba6960d994 | | model-00008-of-00008.safetensors | Weights | 1.1 GB | 1b9d9b328f85 | | config.json | Configuration | 3.5 KB | — | | model.safetensors.index.json | Configuration | 120.2 KB | — | | README.md | Documentation | 4.3 KB | — | | figures/DER.jpg | Other | 62.7 KB | — | | figures/VibeVoice_ASR_archi.png | Other | 149.5 KB | c5973094212c | | figures/cpWER.jpg | Other | 68.5 KB | — | | figures/language_distribution_horizontal.png | Other | 888.4 KB | 1151d7587c66 | | figures/tcpWER.jpg | Other | 64.3 KB | — | | .gitattributes | Repository | 199 B | — | ## License and Download License mit Access Open weights, no gate Download size 17.3 GB [Download from Microsoft](https://huggingface.co/microsoft/VibeVoice-ASR) Released by Microsoft through its official repository on Hugging Face. [Read the license](https://opensource.org/license/mit). ## Built From - Described by arXiv:2601.18184 ## Memory Requirements | Precision | Weights in memory | | --- | --- | | As published | 17.3 GB | | 16-bit | 17.3 GB | | 8-bit | 8.7 GB | | 4-bit | 4.3 GB | Weights only, from the published parameter count; the key-value cache and runtime add to this. ## Built on This Model - Derived from[VibeVoice-ASR-DFlash2](https://savrn.com/models/vibevoice-asr-dflash2) ## Questions About VibeVoice-ASR ### How much GPU memory does VibeVoice-ASR need? About 20.8 GB at 16-bit and 5.2 GB at 4-bit: the weights (8.7B parameters) plus a working margin. A long context needs more. ### What is the cheapest GPU to run VibeVoice-ASR 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 VibeVoice-ASR commercially? Yes. VibeVoice-ASR 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 Model · Speech recognition ### [VibeVoice-ASR-DFlash2](https://savrn.com/models/vibevoice-asr-dflash2) [Nikita Davidchuk](https://savrn.com/model-publishers/ar4ikov) VibeVoice-ASR (7B), the original BF16 weights as Microsoft published them, with its DFlash 2 drafter bundled in drafter/: one download, and vibevoice.c decodes with speculative decoding -- the drafter proposes 8 tokens in one pass, the model checks them in one pass and keeps the ones it agrees with. the model's own decode step, so the transcript is byte-for-byte the one without the drafter. (With a drafter vibevoice.c attends with flashinfer, which checks rows together; the transcript is then --attn flashinfer's.) Needs vibevoice.c with DFlash 2 support: branch dflash2 (PR #48), in the next release. A model directory's drafter/ is used without asking: vibevoice.c runs BF16 weights as dense… Open weights mit 8.7B parameters vibevoice.c [View model](https://savrn.com/models/vibevoice-asr-dflash2) Model · Speech recognition ### [VibeVoice-ASR-AWQ-W4A16-ASYM-DFlash2](https://savrn.com/models/vibevoice-asr-awq-w4a16-asym-dflash2) [Nikita Davidchuk](https://savrn.com/model-publishers/ar4ikov) VibeVoice-ASR (7B), AWQ W4A16 (asymmetric, groups of 128), with its DFlash 2 drafter bundled in drafter/: one download, and vibevoice.c decodes with speculative decoding -- the drafter proposes 8 tokens in one pass, the model checks them in one pass and keeps the ones it agrees with. the model's own decode step, so the transcript is byte-for-byte the one without the drafter. (With a drafter vibevoice.c attends with flashinfer, which checks rows together; the transcript is then --attn flashinfer's.) Needs vibevoice.c with DFlash 2 support: branch dflash2 (PR #48), in the next release. A model directory's drafter/ is used without asking: vibevoice.c b72be15 (branch dflash2), RTX 3090, greedy… Open weights mit 8.3B parameters vibevoice.c [View model](https://savrn.com/models/vibevoice-asr-awq-w4a16-asym-dflash2) Model · Speech recognition ### [Voxtral-Mini-4B-Realtime-2602](https://savrn.com/models/voxtral-mini-4b-realtime-2602) [Mistral AI_](https://savrn.com/model-publishers/mistralai) Voxtral Mini 4B Realtime 2602 is a multilingual, realtime speech-transcription model and among the first open-source solutions to achieve accuracy comparable to offline systems with a delay of = 3600 / 0.8 = 45000. In theory, you should be able to record with no limit; in practice, pre-allocations of RoPE parameters among other things limits --max-model-len. For the best user experience, we recommend to simply instantiate vLLM with the default parameters which will automatically set a maximum model length of 131072 (~ca. 3h). Open weights apache-2.0 4.4B parameters 131,072 tokens vllm [View model](https://savrn.com/models/voxtral-mini-4b-realtime-2602) Model · Speech recognition ### [Audio8-ASR-Infinite](https://savrn.com/models/audio8-asr-infinite) [Edge0](https://savrn.com/model-publishers/edge0) Audio8 ASR Infinite is a native streaming speech recognition model built to be as responsive as possible. It offers a selectable audio clock (80/120/160 ms) and a transcription delay (240–560 ms). With our adapted vLLM build it transcribes unlimited-length audio 24/7 without drifting. - Super responsive — the native streaming architecture decodes 12.5 times per second. - Unlimited-length transcription — a rolling KV Cache keeps both memory and latency constant, even in 24/7 operation. - Selectable streaming clock — one text token per clock step (12.5 / 8.3 / 6.25 decisions per second), balancing perception granularity and resource cost. - Configurable transcription delay — set how much… Open weights apache-2.0 4.1B parameters 32,768 tokens transformers [View model](https://savrn.com/models/audio8-asr-infinite) Model · Speech recognition ### [qwen3-asr-1.7b-int8-mlx](https://savrn.com/models/qwen3-asr-1-7b-int8-mlx) [App Automaton](https://savrn.com/model-publishers/appautomaton) MLX-native int8 conversion of Qwen3-ASR-1.7B for local automatic speech recognition on Apple Silicon — English, Chinese, and mixed Chinese/English. It runs through the mlx-speech runtime with no PyTorch and no cloud API at inference time. Weights ship as plain.safetensors. Install mlx-speech, then load by alias (downloads on first use): Or download the weights once and load by local path: - This repo contains the quantized MLX runtime artifact only. - The quantization mode (affine, groupsize 64, 8-bit) is stored in config.json and re-applied automatically on load, so it cannot desync from the weights. - v0 is single-pass offline ASR; streaming, timestamps, and long-audio chunking are… Open weights apache-2.0 2.4B parameters mlx [View model](https://savrn.com/models/qwen3-asr-1-7b-int8-mlx) Model · Speech recognition ### [Qwen3-ASR-1.7B](https://savrn.com/models/qwen3-asr-1-7b) [Qwen](https://savrn.com/model-publishers/qwen) The Qwen3-ASR family includes Qwen3-ASR-1.7B and Qwen3-ASR-0.6B, which support language identification and ASR for 52 languages and dialects. Both leverage large-scale speech training data and the strong audio understanding capability of their foundation model, Qwen3-Omni. Experiments show that the 1.7B version achieves state-of-the-art performance among open-source ASR models and is competitive with the strongest proprietary commercial APIs. Here are the main features: Novel and strong forced alignment Solution: We introduce Qwen3-ForcedAligner-0.6B, which supports timestamp prediction for arbitrary units within up to 5 minutes of speech in 11 languages. Evaluations show its timestamp… Open weights apache-2.0 2.3B parameters [View model](https://savrn.com/models/qwen3-asr-1-7b) ## Microsoft [All models and datasets](https://savrn.com/model-publishers/microsoft) ## Versions - [d0c9efdb8d61](https://savrn.com/models/vibevoice-asr/versions/d0c9efdb8d61) · current 2026-09-18 ## Explore More - [All speech recognition models](https://savrn.com/models/tasks/speech-recognition) - [All models under mit](https://savrn.com/models/licenses/mit) - [Model comparisons](https://savrn.com/models/comparisons) - [The model directory](https://savrn.com/models) - [Open model prices by host](https://savrn.com/ai-index/pricing/open-models) ## Source - Repository metadata, read 2026-10-19. - [Microsoft website](https://github.com/microsoft) - [Hugging Face record](https://huggingface.co/microsoft/VibeVoice-ASR) - [How the hub is built](https://savrn.com/model-hub/methodology)