Lexus2016/cheap-llm-router

Save Claude/Codex subscription tokens by routing context-only file reads through a cheaper LLM (DeepSeek, Kimi, Gemini Flash). One-line install, works with any OpenAI-compatible model.

Visit GitHub

Python

2 contributors

MIT License

FeaturedAdvertise to 180K+ monthly visitors

Bifrost

Fastest Enterprise AI Gateway for Routing, Governing, and Securing all your AI Traffic

Momentum

How fast Lexus2016/cheap-llm-router 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

Oct 30 – Nov 5, 2026

Stars +1 in 7 days

Forks +1 in 7 days

Merged PRs 0 in 7 days

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

Not in the Trendshift top 25 in this period

Oct 30Oct 31Nov 1Nov 2Nov 3Nov 4Nov 5

Stars, forks, merged pull requests and Trendshift Trending rank, by day
Day	Stars gained	Forks gained
Fri, Oct 30, 2026	+0	+1
Sat, Oct 31, 2026	+0	+0
Sun, Nov 1, 2026	+0	+0
Mon, Nov 2, 2026	+1	+0
Tue, Nov 3, 2026	+0	+0
Wed, Nov 4, 2026	+0	+0
Thu, Nov 5, 2026	+0	+0
Total	+1	+1

#1 on the list
Rank that day
Not trending

Social mentions

Recent discussions about this repository across the web

Nov 51 post
- Telling Haiku 5.5 to "answer directly" does not stop it thinking. Anthropic says so in its own prompting guide. Thinking is on by default and counts toward max_tokens. Going from low to medium effort…
  Read full on X, posted by @YevMur
  @YevMur · x.com
Nov 11 post
- 11 a.m. and my AI says come back later. The limit wasn't eaten by thinking. It was eaten by reading whole files. What fixed it: 1. Point it at the exact file and function. 2. One task, one chat. 3.…
  Read full on X, posted by @YevMur
  @YevMur · x.com
Oct 301 post
- OpenAI was going to halve ChatGPT Pro limits today. After the backlash, they doubled them. Fine print: "A larger allowance does not guarantee twice as many completed tasks." How I stretch mine: 1.…
  Read full on X, posted by @YevMur
  @YevMur · x.com
Oct 291 post
- GitHub shipped Copilot Fusion yesterday, and the clever bit isn't picking the best model. It's "Cascade": a cheap model drafts, a check decides, and only a fail goes up to the expensive one. I do the…
  Read full on X, posted by @YevMur
  @YevMur · x.com
1 month earlier
Sep 231 post
- open-sourced Unreal Agent (Sep 22): async tool-call harness so the model stops paying tokens on waits, polls, and heartbeats. Up to 40% cost vs Codex on Terminal-Bench / SWE, similar pass rates.
  Read full on X, posted by @YevMur
  @YevMur · x.com
Sep 202 posts
- What I ship in public — local agent tooling from the ugly middle: LocalGuard · cheap-llm-router · claude-code-studio · LLM-Fusion Links in the next reply.
  Read full on X, posted by @YevMur
  @YevMur · x.com
- shipped a smart router cutting coding-agent spend ~60%: planning on frontier models, 84% of edits on cheaper open weights (Mirror benchmarking on live sessions). Same hole I hit. Agents burn Opus…
  Read full on X, posted by @YevMur
  @YevMur · x.com
Sep 183 posts
- CERA-MoA just dropped on @arxiv (Sep 16): Mixture-of-Agents where the router and the agents co-evolve, instead of a frozen routing map next to static workers. Same problem I keep hitting in…
  Read full on X, posted by @YevMur
  @YevMur · x.com
- Fuse V2 just mashed complexity + capability scoring on SWE-bench Verified: 25/25 solved at $0.75 per solve.
  Read full on X, posted by @YevMur
  @YevMur · x.com
- on "Your Agent Is Mine": of 428 LLM routers measured, 9 injected code and 17 abused credentials. A router can rewrite tool calls after the model answers.
  Read full on X, posted by @YevMur
  @YevMur · x.com