salesforce-sovereign-intelligence/SOV-Transcribe-1
just landedmodel
SOV-Transcribe-1 is Salesforce's open-weight speech-augmented language model built for non-streaming automatic speech recognition (ASR). Mixture-of-experts, 27B total, 4.6B active (128 experts, 8 per token). Context: 256K tokens (config). License: apache-2.0. Vendor-reported: "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."
- Vendor-reported: 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.
- Org
- salesforce-sovereign-intelligence
- Released
- 2026-11-10
- Status
- just landed
- License
- apache-2.0 (permissive)
- Parameters
- 26.6B total, 4.6B active
- Architecture
- MoE · 128 experts, 8/token · SovereignAITranscribeForConditionalGeneration
- Context
- 262,144 tokens (config.json)
- Task
- automatic-speech-recognition · audio
- Signal
- 56 HF likes · 463 downloads
Runs on
mixed: real files + estimate; need = weights + FP16 KV cache for 8K tokens + 1GB. A sizing calculation, not tested.
- Q4 ~19.0 GB (estimated size): 24GB GPU (3090 / 4090), 32GB GPU (5090), 24GB GPU (7900 XTX), Mac 36GB, CPU only, 32GB RAM (slow)
- Q8 ~31.3 GB (estimated size): 48GB GPU (RTX 6000 Ada / A6000), Mac 48GB, Mac 64GB (M1 Max), CPU only, 64GB RAM (slow)
- FP16 ~56.3 GB: 96GB GPU (RTX Pro 6000 Blackwell), 1× DGX Spark (128GB), 2× DGX Spark (256GB), Strix Halo 128GB unified (~96GB usable), Mac 96GB, Mac 128GB, Mac 192GB, Mac 256GB, Mac 512GB, CPU only, 128GB RAM (slow), CPU only, 256GB RAM (slow)
Links
First published here 2026-11-10 · updated 2026-11-13 · markdown · JSON