0xBakeer/TandemLLM — GitHub trending stats & insights | Trendshift

0xBakeer/TandemLLM

Inference engine for Qwen3.8-27B on one DGX Spark: speculative draft trees sized by StairCut, NVFP4 kernels, exact recurrent-state caches

#AI infrastructure

Python

33

2

1 contributors

GNU Affero General Public License v3.0

FeaturedAdvertise to 180K+ monthly visitors

Enterprise AI Gateway Built For Scale: Route, Govern, and Secure All Your AI Traffic

Control what AI can do before it acts

Run LLMs locally in Flutter, React Native, Swift, Kotlin, Python and Godot

Your agents build it. Kane verifies it. Writes the tests, runs them, and maintains them

Momentum

How fast 0xBakeer/TandemLLM is gaining stars, and where it lands on Trendshift and GitHub Trending

Activity and trending

Stars, forks and merged pull requests each day or month, lined up with the repository's rank on Trendshift Trending

Sep 28 – Oct 13, 2026

Stars +33 in 16 days

11

Forks +2 in 16 days

1

Merged PRs 0 in 16 days

Trendshift Trendingwhere the repository ranked each day it made the top 25

Not in the Trendshift top 25 in this period

Sep 28Sep 30Oct 2Oct 4Oct 6Oct 8Oct 10Oct 12

Stars, forks, merged pull requests and Trendshift Trending rank, by day
Day	Stars gained	Forks gained	Merged PRs
Mon, Sep 28, 2026	+0	+0	0
Tue, Sep 29, 2026	No figure	No figure	No figure
Wed, Sep 30, 2026	+11	+0	0
Thu, Oct 1, 2026	+6	+0	0
Fri, Oct 2, 2026	+3	+0	0
Sat, Oct 3, 2026	+2	+1	0
Sun, Oct 4, 2026	+1	+0	0
Mon, Oct 5, 2026	+1	+0	0
Tue, Oct 6, 2026	+3	+0	0
Wed, Oct 7, 2026	No figure	No figure	No figure
Thu, Oct 8, 2026	No figure	No figure	No figure
Fri, Oct 9, 2026	+1	+1	0
Sat, Oct 10, 2026	+1	+0	0
Sun, Oct 11, 2026	+2	+0	0
Mon, Oct 12, 2026	No figure	No figure	No figure
Tue, Oct 13, 2026	+2	+0	0
Total	+33	+2	0

#1 on the list
Rank that day
Not trending

Social mentions

Recent discussions about this repository across the web

Oct 131 post
- Meet Wren, the small model in TandemLLM's setup page. It runs on your own box and walks you through installing a model, one step at a time. But picking weights for a box is not obvious. Wren asks…
  Read full on X, posted by @0xBakeer
  @0xBakeer · x.com
12 days earlier
Oct 13 posts
- The repo describes the lookup drafter and the suffix store in detail: docs/speculative-decoding.md: the three sources, 0.2 ms against 25 ms, the deep chain (12.6 → 15.75 tokens per round), and the…
  Read full on X, posted by @0xBakeer
  @0xBakeer · x.com
- learned n-gram embedding tables, both in DeepSeek-V4.1-Flash and Qwen3.8-Flash-Next
  Read full on X, posted by @0xBakeer
  @0xBakeer · x.com
- Speculative decoding has one knob most engines fix: how many guessed tokens to check per step. At 1k context, checking 16 guessed tokens costs TandemLLM engine the same 74 ms as checking 2. At 17 it…
  Read full on X, posted by @0xBakeer
  @0xBakeer · x.com
Sep 291 post
- One line installs it on a Spark: curl -fsSL | bash It checks the GPU, downloads the base model, the NVFP4 weights and both drafters, then starts an OpenAI-compatible server with a dashboard. I only…
  Read full on X, posted by @0xBakeer
  @0xBakeer · x.com