# Blog

> LightSeek Foundation Blog: research notes and engineering updates on open-source AI systems.

## Table of Contents

- [EAGLE 3.1: Advancing Speculative Decoding Through Collaboration Between the EAGLE Team, vLLM, and TorchSpec](/blog/eagle-3-1.md)
- [Kimi K3 at Day 0: Frontier Model Enablement on Leading Platforms with TokenSpeed](/blog/tokenspeed-kimi-k3.md)
- [Kimi K3 Optimization on GB300 — Part I](/blog/kimi-k3-optimization-gb300-part-i.md)
- [Qwen3.8 at Day 0: 2.4T-Scale Inference with TokenSpeed](/blog/tokenspeed-qwen3-8.md)
- [Releasing Kimi K3 Draft Collection: Efficient Draft Training in TorchSpec](/blog/kimi-k3-draft-collection.md)
- [TML Inkling at Day 0: FP4 Inference on NVIDIA and AMD with TokenSpeed](/blog/tokenspeed-inkling.md)
- [TokenSpeed: A Speed-of-Light LLM Inference Engine for Agentic Workloads](/blog/lightseek-tokenspeed.md)
- [TorchSpec: Speculative Decoding Training at Scale](/blog/torchspec-speculative-decoding-training-at-scale.md)