RadixArk
/

Qwen3.8-Flash-Next-NVFP4

Image-Text-to-Text

Model Optimizer

8-bit precision

Model card Files Files and versions

Resources

View closed (0)

# [Bug] NEXTN/MTP draft acceptance decays to ~0 over server uptime on qwen4_exp (Qwen3.8-Flash-Next), fully restored by a restart

#8 opened about 3 hours ago by

One-command deployment on a single DGX Spark (34-42 tok/s, prefix caching, 262K)

#7 opened 3 days ago by

MTP block is unquantized (4.86 GiB BF16) — makes speculative decoding a net LOSS on pre-Hopper GPUs (fix: 1.5 GiB transplant, 14 -> 52 tok/s)

#6 opened 3 days ago by

Optimized SGLang for Qwen3.8 Flash-Next NVFP4 on 1x RTX PRO 6000: 171 tok/s, 524K, and HiCache/NIXL

#5 opened 4 days ago by

Field notes from a 2x DGX Spark (GB10) deployment of this quant — wedge recovery, firewall deadlock, MTP acceptance data

#4 opened 4 days ago by

Degeneration Loop when calling tool (PDF/Image)

#3 opened 5 days ago by

Startup issues with vllm-openai:qwen38-flash-next

#2 opened 5 days ago by

Tooooo BIG for a single node DGX Spark :(

#1 opened 6 days ago by