Comparison market
Speech, audio, and music
Every listed pair shares this sourced market, passes the comparison evidence gate, and uses one canonical unordered URL.
28 published comparisons
Primary-source records
vs | |||||
|---|---|---|---|---|---|
 Granite Speech 5.0 TurboCTCibm-granite/granite-speech-5.0-470m-turboctc | vs |  VibeVoice ASR Streaming 1.5Bmicrosoft/VibeVoice-ASR-Streaming-1.5B | IBM / Microsoft | audio, text | cross-developer peers |
 Lyria 3.5lyria-3.5 | vs |  Stable Audio 3 Mediumstabilityai/stable-audio-3-medium | Google DeepMind / Stability AI | audio | cross-developer peers |
 MiniMax Music3MiniMaxAI/MiniMax-Music3 | vs | Lyria 3.5lyria-3.5 | MiniMax / Google DeepMind | audio | cross-developer peers |
Seed 2.0 Litebytedance-seed/seed-2.0-lite | vs |  Inklingthinkingmachines/Inkling | ByteDance Seed / Thinking Machines Lab | audio, image, text, video | cross-developer peers |
 Muse Spark 1.1meta-llama/muse-spark-1.1 | vs | Inklingthinkingmachines/Inkling | Meta / Thinking Machines Lab | audio, image, text, video | cross-developer peers |
Seed 2.0 Litebytedance-seed/seed-2.0-lite | vs | Muse Spark 1.1meta-llama/muse-spark-1.1 | ByteDance Seed / Meta | audio, image, text, video | cross-developer peers |
Seed 2.0 Litebytedance-seed/seed-2.0-lite | vs | Muse Spark 1.2meta-llama/muse-spark-1.2 | ByteDance Seed / Meta | audio, image, text, video | cross-developer peers |
Gemma 4 E2Bgoogle/gemma-4-E2B-it | vs | Muse Spark 1.1meta-llama/muse-spark-1.1 | Google DeepMind / Meta | audio, image, text, video | cross-developer peers |
Gemma 4 E4Bgoogle/gemma-4-E4B-it | vs | Muse Spark 1.1meta-llama/muse-spark-1.1 | Google DeepMind / Meta | audio, image, text, video | cross-developer peers |
Gemma 4 12Bgoogle/gemma-4-12B-it | vs | Muse Spark 1.2meta-llama/muse-spark-1.2 | Google DeepMind / Meta | audio, image, text, video | cross-developer peers |
Gemma 4 E4Bgoogle/gemma-4-E4B-it | vs | Muse Spark 1.2meta-llama/muse-spark-1.2 | Google DeepMind / Meta | audio, image, text, video | cross-developer peers |
Lyria 3.5lyria-3.5 | vs | Inklingthinkingmachines/Inkling | Google DeepMind / Thinking Machines Lab | audio, image, text | cross-developer peers |
Lyria 3.5lyria-3.5 | vs | Phi-4 Multimodal Instructmicrosoft/Phi-4-multimodal-instruct | Google DeepMind / Microsoft | audio, image, text | cross-developer peers |
Lyria 3.5lyria-3.5 | vs | Muse Spark 1.1meta-llama/muse-spark-1.1 | Google DeepMind / Meta | audio, image, text | cross-developer peers |
Granite Speech 5.0 TurboCTC NCibm-granite/granite-speech-5.0-470m-turboctc-nc | vs | VibeVoice ASR Streaming 1.5Bmicrosoft/VibeVoice-ASR-Streaming-1.5B | IBM / Microsoft | audio, text | cross-developer peers |
Granite Speech 5.0 TurboCTCibm-granite/granite-speech-5.0-470m-turboctc | vs | VibeVoice ASR Streaming 7Bmicrosoft/VibeVoice-ASR-Streaming-7B | IBM / Microsoft | audio, text | cross-developer peers |
Granite Speech 5.0 TurboCTC NCibm-granite/granite-speech-5.0-470m-turboctc-nc | vs | VibeVoice ASR Streaming 7Bmicrosoft/VibeVoice-ASR-Streaming-7B | IBM / Microsoft | audio, text | cross-developer peers |
Granite Speech 5.0 TurboCTCibm-granite/granite-speech-5.0-470m-turboctc | vs | Muse Spark 1.2meta-llama/muse-spark-1.2 | IBM / Meta | audio, text | cross-developer peers |
Granite Speech 5.0 TurboCTC NCibm-granite/granite-speech-5.0-470m-turboctc-nc | vs | Muse Spark 1.2meta-llama/muse-spark-1.2 | IBM / Meta | audio, text | cross-developer peers |
Muse Spark 1.2meta-llama/muse-spark-1.2 | vs | VibeVoice ASR Streaming 1.5Bmicrosoft/VibeVoice-ASR-Streaming-1.5B | Meta / Microsoft | audio, text | cross-developer peers |
MiniMax Music3MiniMaxAI/MiniMax-Music3 | vs | Stable Audio 3 Mediumstabilityai/stable-audio-3-medium | MiniMax / Stability AI | audio | cross-developer peers |
VibeVoice ASR Streaming 1.5Bmicrosoft/VibeVoice-ASR-Streaming-1.5B | vs | VibeVoice ASR Streaming 7Bmicrosoft/VibeVoice-ASR-Streaming-7B | Microsoft / Microsoft | audio, text | family variants |
Phi-4 Multimodal Instructmicrosoft/Phi-4-multimodal-instruct | vs | VibeVoice ASR Streaming 1.5Bmicrosoft/VibeVoice-ASR-Streaming-1.5B | Microsoft / Microsoft | audio, text | developer peers |
Phi-4 Multimodal Instructmicrosoft/Phi-4-multimodal-instruct | vs | VibeVoice ASR Streaming 7Bmicrosoft/VibeVoice-ASR-Streaming-7B | Microsoft / Microsoft | audio, text | developer peers |
Granite Speech 5.0 TurboCTCibm-granite/granite-speech-5.0-470m-turboctc | vs | Granite Speech 5.0 TurboCTC NCibm-granite/granite-speech-5.0-470m-turboctc-nc | IBM / IBM | audio, text | family variants |
Inklingthinkingmachines/Inkling | vs | Inkling Smallthinkingmachines/Inkling-Small | Thinking Machines Lab / Thinking Machines Lab | audio, image, text | family variants |
Gemma 4 12Bgoogle/gemma-4-12B-it | vs | Gemma 4 E2Bgoogle/gemma-4-E2B-it | Google DeepMind / Google DeepMind | audio, image, text, video | family variants |
Gemma 4 12Bgoogle/gemma-4-12B-it | vs | Gemma 4 E4Bgoogle/gemma-4-E4B-it | Google DeepMind / Google DeepMind | audio, image, text, video | family variants |
Reviewed pairs are shown by default and included in the comparison sitemap. Model A and Model B searches also cover other published pairs in this market.
Questions
Speech, audio, and music Comparison FAQs
What is included in Speech, audio, and music model comparisons?+
This market currently contains 28 published speech, audio, and music comparisons. Every pair shares a sourced category or workload and has enough compatible fields to support a direct page.
How are Speech, audio, and music comparison pairs selected?+
Pairs are generated from compatible sourced records, then curated when they represent useful family variants, developer peers, or cross-developer peers. Curation controls discovery priority; it does not determine a winner.
Does the order of two models change a comparison?+
No. Every pair has one canonical unordered URL, so reversing the model names resolves to the same comparison and the same underlying evidence.
Why is a speech, audio, and music model pair missing?+
A pair remains unpublished when the models do not share the required sourced category or tasks, when too few fields are comparable, or when one record lacks enough primary evidence. Missing values are not invented to make a pair eligible.
Which speech, audio, and music model is best?+
There is no category-wide winner. Individual comparison pages identify field-level differences and protocol-matched benchmark leaders when available; price, deployment, context, task fit, and evidence gaps remain separate decisions.
How current are these speech, audio, and music comparisons?+
The comparison directory was last refreshed Oct 19, 2026. Each detail page retains its underlying source links and observation dates so later model changes are not presented as already verified.