DFlash 2: 20% More Tokens from Every Verification Pass
DFlash 2: Keep Drafting Parallel
Inco AI's DFlash 2 upgrades speculative decoding by adding a lightweight path selector and a convolutional fix for suffix decay, boosting output by 16–25% with only ~1% added latency. The selector scores adjacent candidate pairs to pick a coherent path, while convolutions target within-block dependencies that attention neglects. Already integrated into SGLang, vLLM, llama.cpp, and oMLX, DFlash 2 serves Qwen3.8-27B at 2.7–3.4× the throughput of autoregressive decoding.
Choosing is cheaper than predicting.
- ilc
Watch the video carefully. DFlash2's tool call fails on python syntax.
Usually models in this class nail things like that 1 shot, which the other side did.
I don't know the cause. It may be nothing. But I'd like to see the model doing something where its path is a bit more constrained, to help out rule out such oddities.
- hypfer
Amazing tech
> An agent writes in an afternoon what a chatbot writes in a month
But can you just.. not.
Your tech is so good, it speaks for itself. Don't ruin that.
- adefa
I'm getting around 27 tokens per second decode using vLLM + Qwen 3.8 27b nvfp4 + DFlash 2 on the DGX Spark.