qwen 4 exp is also the architecture name for qwen 3.8 flash next so maybe

Try my dflash2 head huggingface.co/ProCreations/…

Nanbeige4.3-33B DFlash2 speculative drafter ⚡️ 1.76× faster decoding on BF16, 1.65× on FP8, 1.51× on NVFP4. Agentic coding ~2.6×. Same outputs: exact rejection sampling. This is an early result, an improved version with higher acceptance is in training. SGLang 0.5.21, drop-in: huggingface.co/ProCreations/…

Yeaaaa it’s annoying… full context can use 100gb+ bf16 kv cache

my dflash2 head finishes soon!

training something awesome overnight alert 🚨 two things actually, second one needs more time

nothing significant

i love it when the best open weight model in the world (kimi k3.1 pro) randomly starts speaking another language and ramble about mouses and hospitals in the middle of a security audit

Kimi K3.1 Pro (max) on black hole benchmark 1.0 (2.0, which is much harder, is running now) This isn’t bad, it isn’t great. Its just “good”. Below sonnet 5.5 for sure I’d put it in between GPT 6.1 Sol and GLM 5.3

I think openai gotta start distilling off of kimi or their cooked

no “preview” at the end 👀 hopefully their further RL cooked

you guys have always been one of the best open weight labs in my opinion, please open weight this

ty!!

As you post this, I got accepted :)

2 million downloads

danggg you got the version straight out of their lab

try @AsideAI

plush :D

who is railway, why is everyone getting hacked, and why won't anthropic let me into Duet