DSpark
by DeepSeek · Technique · Text & code
Multi-token-per-step decoding component included in V4.1 Flash
Evidence
- Sep 7DSpark gave 1.15–1.63x on Qwen3-8B on AMD GPUs versus 1.08–1.27x for DFlash
- Aug 16Speculative decoding technique added to DeepSeek V4 Pro 0813
- Aug 2Lucebox's DSpark drafter uses DeepSeek V4 Flash's built-in multi-token-prediction heads, with no separate draft model needed