salesforce-sovereign-intelligence/SOV-Transcribe-1 · Hugging Face
retained text
# SOV-Transcribe-1 SOV-Transcribe-1 is Salesforce's open-weight speech-augmented language model built for non-streaming automatic speech recognition (ASR). It combines leading English transcription accuracy with multilingual support and natural-language control over how speech is transcribed. The model ranks **#1 among open-weight models for non-streaming speech-to-text on Artificial Analysis (#7 overall)** and **#2 of 23 systems on VoiceCodeBench ( besimple AI )** for entity accuracy and task success as of November 12th, 2026. Developers and enterprises can deploy and adapt SOV-Transcribe-1 on their own infrastructure, retaining control over their audio data and transcription behavior. # [](https://huggingface.co/salesforce-sovereign-intelligence/SOV-Transcribe-1#key-features)Key features * **Accurate entity recognition.** Captures details such as names, phone numbers and email addresses that matter in dictation and voice applications. * **Robustness to accents and noise.** Handles diverse English accents and challenging audio conditions, including background noise. * **Transcription style control.** Use natural-language instructions to choose verbatim or cleaned-up transcripts and control punctuation, capitalization and number formatting. * **Context biasing.** Supply names, domain-specific vocabulary or a short description of the topic to help recognize difficult words and technical terms. # [](https://huggingface.co/salesforce-sovereign-intelligence/SOV-Transcribe-1#model-summary)Model summary |Property|Details| |:-|:-| |**Total parameters**|26.6B total (4.6B active) including the audio encoder| |**Architecture**|Whisper large-v3 audio encoder, audio projector and Gemma 4 26B A4B MoE| |**Precision**|BF16 (MoE router in FP32)| |**Maximum audio input**|Up to 30 minutes (tested)| |**Inputs / outputs**|16 kHz Audio with optional text instructions and context / transcript text| |**Languages**|English, French, German, Spanish, Italian, Portuguese and Dutch| |**Tasks**|Verbatim or written-form transcription, multilingual ASR, style control and contextual transcription| |**Inference frameworks**|vLLM and Hugging Face Transformers| |**Speculative decoding**|DFlash draft model in its own repository, [`salesforce-sovereign-intelligence/SOV-Transcribe-1-DFlash`](https://huggingface.co/salesforce-sovereign-intelligence/SOV-Transcribe-1-DFlash)| |**License**|Apache 2.0| # [](https://huggingface.co/salesforce-sovereign-intelligence/SOV-Transcribe-1#benchmark-results) * **DFlash** : [https://huggingface.co/salesforce-sovereign-intelligence/SOV-Transcribe-1-DFlash](https://huggingface.co/salesforce-sovereign-intelligence/SOV-Transcribe-1-DFlash) * **Space** : [https://huggingface.co/spaces/hugging-apps/sov-transcribe-1](https://huggingface.co/spaces/hugging-apps/sov-transcribe-1) * **Blog Post** : [https://huggingface.co/blog/salesforce-sovereign-intelligence/sov-transcribe-1](https://huggingface.co/blog/salesforce-sovereign-intelligence/sov-transcribe-1)
document fetched 2026-11-15 00:36 · canonical https://reddit.com/r/LocalLLaMA/comments/1xvst4u/salesforcesovereignintelligencesovtranscribe1