# lightseekorg/tokenspeed

> TokenSpeed is an open-source LLM inference engine for agentic workloads, combining TensorRT-LLM-level performance with vLLM-level usability. It separates a C++ control plane from a Python execution plane and ships a pluggable kernel subsystem including fast MLA implementations.

- Magnitude: 3.4 out of 10 — Breakout
- Stars: 2,310 total · +33 stars measured 2026-10-15, 11:22–13:00 UTC, ≈ 54 by evening
- Star trust: star growth looks organic
- Category: Language models · Language: Python · License: MIT · Created: 2026-05-06 · Last push: 2026-10-15
- GitHub: https://github.com/lightseekorg/tokenspeed · Homepage: https://lightseek.org/blog/lightseek-tokenspeed.html · Page: https://gitnova.dev/en/r/lightseekorg/tokenspeed

## Useful for

- Deploy a low-latency LLM server for agentic workloads on NVIDIA or AMD GPUs
- Benchmark tokens-per-second per user against TensorRT-LLM on the same model and hardware
- Extend the kernel registry with custom kernels for a new accelerator backend

## Why it’s here

- Star-counter measurements on 2026-10-15 (UTC), 11:22–13:00: 2277 → 2310 stars (+33). This is the change over that interval.
- Estimated end-of-day forecast: about +54 stars, using observed gains and the previous day.
- Before this, the repository barely got any stars — about 2 per day.
- Elevated gains were measured on the last 2 completed UTC days in a row. Continued growth is not guaranteed.
- GitHub Trending today: #11, +25 stars.
- GitHub Trending Python today: #4, +25 stars.

## Star trust

Star growth looks organic.

Star-trust labels are heuristics based on the repository’s behavior, not a check of every stargazer.

## Numbers

- Forks: 331
- Issues and pull requests: 2,334
- Watchers: 14
- Average over the last week: 16 per day
- Usual pace: 3 per day
- Stars in the last hour (measured): 20
- Latest release: v0.1.1 (2026-10-06)

## Stars per day, last 30 days (oldest → newest, today is partial)

2026-09-16 … 2026-10-15: 5, 5, 3, 3, 2, 5, 7, 2, 5, 4, 3, 4, 2, 3, 1, 2, 1, 1, 0, 0, 1, 7, 5, 3, 3, 2, 3, 18, 31, 33

## Spotted in now

- GitHub Trending today: #11, +25 stars
- GitHub Trending Python today: #4, +25 stars

## More in this category

1. **chenglou/handwired** — 7.1 · Early signal · Language models · TypeScript · +248 stars measured 2026-10-15, 06:44–12:56 UTC, ≈ 359 by evening
   Turns regular JavaScript code into a differentiable neural net: if-else, loops and state are compiled into matrix multiplications and ReLU with no training. Runs on JS or WebGPU.
   Full card: https://gitnova.dev/en/r/chenglou/handwired.md
2. **BaruqueRodrigues/llm-legislativo** — 5.7 · Early signal · Language models · Python · +42 stars measured 2026-10-15, 00:23–12:56 UTC, ≈ 96 by evening
   Large-scale Brazilian legislative corpus and a causal Transformer language model trained from scratch in pure PyTorch without pretrained weights or closed APIs.
   Full card: https://gitnova.dev/en/r/BaruqueRodrigues/llm-legislativo.md
3. **prabanta-dev/janas** — 5.0 · Early signal · Language models · C · +14 stars measured 2026-10-15, 00:20–12:56 UTC, ≈ 32 by evening
   Janas is a family of AI tools in C with no dependencies. Janas-LLM runs large MoE models (up to 80B, 48 GB) on a 32 GB laptop by streaming experts from NVMe SSD, and outperforms llama.cpp on all 13 tested models.
   Full card: https://gitnova.dev/en/r/prabanta-dev/janas.md
4. **Niko1221/Strata** — 4.6 · Steady · Language models · C++ · +510 stars measured 2026-10-15, 00:21–12:57 UTC, ≈ 916 by evening
   A C++ local inference engine that runs the 125B MoE model Qwen3.8-Flash-Next on a regular PC with one NVIDIA GPU (12-24 GB) and 64 GB RAM, exposing an OpenAI/Anthropic-compatible API on localhost.
   Full card: https://gitnova.dev/en/r/Niko1221/Strata.md
5. **cdcailf/cjkfold** — 4.6 · Early signal · Language models · Python · +36 stars measured 2026-10-15, 00:24–12:57 UTC, ≈ 70 by evening
   Python library that folds look-alike CJK glyphs (Kangxi radicals and CJK radical supplements) back to the Han characters they render as, so text extracted from PDFs matches in search and comparison. Zero-dependency, pure functions.
   Full card: https://gitnova.dev/en/r/cdcailf/cjkfold.md

---

Magnitude (0–10) measures how fast and how unusually interest in a repository is growing right now. It is not a quality score.
Days are UTC. “So far today” is a fact; “expected by the end of the day” is a forecast.
Summaries and use cases are written by an LLM (DeepSeek V4.1 Flash) from the README and may be inaccurate: verify specific claims (benchmarks, speed, hardware) in the repository itself.
Data as of 2026-10-15 13:16 UTC, updated every 30 minutes.