autotrust/Qwen3.8-Flash-Next-E384-FP4-V2
just landedmodel
Image text to text model from autotrust (HF pipeline tag; the card has no one-line description). Mixture-of-experts, 80B total (384 experts, 10 per token). Context: 256K tokens (config). License: nvidia-open-model-agreement.
- Org
- autotrust
- Released
- 2026-10-26
- Status
- just landed
- License
- nvidia-open-model-agreement (restricted)
- Parameters
- 80.1B
- Architecture
- MoE · 384 experts, 10/token · Qwen4ExpForConditionalGeneration
- Context
- 262,144 tokens
- Task
- image-text-to-text · vision
- Signal
- 243 HF likes · 2386 downloads
Runs on
estimate from params; need = weights + FP16 KV cache for 8K tokens + 1GB. A sizing calculation, not tested.
- Q4 ~49.8 GB (estimated size): Mac 96GB, CPU only, 64GB RAM (slow)
- Q8 ~86.9 GB (estimated size): 96GB GPU (RTX Pro 6000 Blackwell), 1× DGX Spark (128GB), Strix Halo 128GB unified (~96GB usable), Mac 128GB, Mac 192GB, CPU only, 128GB RAM (slow)
- FP16 ~162.0 GB (estimated size): 2× DGX Spark (256GB), Mac 256GB, Mac 512GB, CPU only, 256GB RAM (slow)
Links
First published here 2026-10-26 · updated 2026-10-26 · markdown · JSON