adventists-ai/Sophon-Realtime-32B-E2E
archived · low-signalmodel
Sophon-Realtime is a full-duplex voice model: a 32B brain listens to you, decides when you have finished, and writes a reply with sound tags. 32B params. License: apache-2.0.
- Org
- adventists-ai
- Released
- 2026-11-08
- Status
- archived · low-signal
- License
- apache-2.0 (permissive)
- Parameters
- 32.0B
- Architecture
- SophonRealtimeForConditionalGeneration
- Task
- any-to-any · omni
- Signal
- 6 HF likes · 233 downloads
- Archived
- 2026-11-12 (low-signal: aged out below threshold; stats as last seen)
Runs on
estimate from params; need = weights + FP16 KV cache for 8K tokens + 1GB. A sizing calculation, not tested.
- Q4 ~22.9 GB (estimated size): 24GB GPU (3090 / 4090), 32GB GPU (5090), 24GB GPU (7900 XTX), Mac 36GB, Mac 48GB, CPU only, 32GB RAM (slow)
- Q8 ~37.7 GB (estimated size): 48GB GPU (RTX 6000 Ada / A6000), Mac 64GB (M1 Max), CPU only, 64GB RAM (slow)
- FP16 ~67.7 GB (estimated size): 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 2026-11-08 · updated 2026-11-15 · markdown · JSON