# vLLM: what I've logged

> vLLM comes up in 2 posts of my AI log, from October 2026 to October 2026, and each item below links to the primary source. The releases I logged include vLLM v0.32.0rc2 and vLLM v0.31.0, filed under open weights, agents and coding and frontier.

## Items

1. **TokenSpeed** (lightseekorg, 13 Oct 2026): An MIT-licensed inference engine aimed at agentic workloads, pitched as TensorRT-LLM-level performance with vLLM-level usability, with separate AMD, NPU and MLA kernel packages. [Primary source](https://github.com/lightseekorg/tokenspeed) · [Post](https://heyaliasif.com/ai/2026-10-15-liquid-lfm25-tokenspeed-llamacpp#item-2)
2. **vLLM v0.32.0rc2** (vLLM, 13 Oct 2026): A release candidate whose notes show a dependency update (Humming, NVFP4 scale workaround removed) in rc1 and a DeepSeek-V4.1-Flash-NVFP4 eval pin in rc2. [Primary source](https://github.com/vllm-project/vllm/releases/tag/v0.32.0rc2) · [Post](https://heyaliasif.com/ai/2026-10-15-liquid-lfm25-tokenspeed-llamacpp#item-4)
3. **vLLM v0.31.0** (vLLM, 12 Oct 2026): 717 commits from 307 contributors, with a new preload command that keeps weights resident across engine restarts. [Primary source](https://github.com/vllm-project/vllm/releases/tag/v0.31.0) · [Post](https://heyaliasif.com/ai/2026-10-13-muse-wonder-self-replication-radar#item-5)

## Posts

- 2026-10-15 [Liquid's 24B on-device agent model, TokenSpeed, llama.cpp updates](https://heyaliasif.com/ai/2026-10-15-liquid-lfm25-tokenspeed-llamacpp)
- 2026-10-13 [Muse Wonder ships without scores, Anthropic tests self-copying](https://heyaliasif.com/ai/2026-10-13-muse-wonder-self-replication-radar)

Canonical: https://heyaliasif.com/ai/topics/vllm