Kimi-K3-DSpark is a custom variant of the Qwen3 model optimized with speculative decoding (DSpark) for faster text generation. It supports Transformers, vLLM, and SGLang backends. The model is designed for high-throughput inference and is available with custom code on Hugging Face.
Kimi K3 DSpark is a Foundation models & chat project. It focuses on accelerating inference speed for large language models using speculative decoding techniques. Kimi K3 DSpark is an open-source project aimed at developers. Kimi K3 DSpark is open source under the MIT license. It runs on the web, the command line, and API.
It is developed by RadixArk, and it first shipped in 2025. Development happens publicly on GitHub with 1k stars and 388 commits in the last 90 days. Key capabilities include text generation, speculative decoding, and vLLM support.
Summary written by a language model from the project’s public pages.
What PulseGate has recorded for this listing
Closest matches by what these projects do