diff --git a/README.md b/README.md index 2f56879..2ecad6a 100644 --- a/README.md +++ b/README.md @@ -69,6 +69,7 @@ Plugins that provide inference optimizations, speculative decoding, parallelism | Plugin | Name | Repository | Notes | | ------ | ---- | ---------- | ----- | | ❄️ **Arctic Inference** | arctic-inference | [![GitHub](https://img.shields.io/badge/GitHub-repo-black?logo=github)](https://github.com/snowflakedb/ArcticInference) | Snowflake's comprehensive optimization suite: Arctic Ulysses (sequence parallelism), Shift Parallelism, speculative decoding (Arctic Speculator, Suffix Decoding), SwiftKV, and 16x faster embeddings. Achieves 3.4x faster request completion. | +| ⛏️ **MineDraft** | minedraft | [![GitHub](https://img.shields.io/badge/GitHub-repo-black?logo=github)](https://github.com/electron-shaders/MineDraft) | Batch parallel speculative decoding plugin that overlaps drafting with verification by maintaining two batches of requests, hiding draft latency behind verification compute. Up to +75% throughput, -39% latency on Qwen3, Llama-3.3, and EAGLE across ShareGPT/LMSYS Arena/Spec-Bench. ([arXiv:2603.18016](https://arxiv.org/abs/2603.18016)) |