LightOn's open 2B code-search agent cuts Claude Code tokens by 34% and cost by 40%
lmoroney · x · 2026-10-13
LightOn released ColGrep-Agent-2B, an open 2.5GB locally-run model that takes file-finding off frontier coding agents' plates. It sits between Claude Code and the repo: searching with multi-vector semantic ColGrep, reading promising files, following references, and returning paths, line ranges and code.
Measured on 52 code-search questions across 13 unseen repos:
- 34% fewer API tokens vs grep delegation
- 40% lower cache-aware API cost
- No statistically significant accuracy change
The insight: agents burn multiple turns grepping and opening files, each turn routed through the frontier model with a growing context. Offloading search to a small local model cuts that. Ships with paper, GitHub, HF model and GGUF, plus a full setup-to-integration tutorial.
More from coding & agent
- FactoryAI ships Environments, cloud sandbox for agents, now GA — matanSF · 2026-10-13
- Researcher on doing science in the agent era: ask the hard questions, verify every line — CatAstro_Piyush · 2026-10-13
- Vibe-coded port of a language itself cuts typechecking time 3x with Opus — maxleiter · 2026-10-13
- Network engineer of 20 years: once you go declarative for AI automation, you never go back — tristanbob · 2026-10-13
- Muse Wonder scores ~42% on senior engineer benchmark vs Sol 6.1's 71% — danshipper · 2026-10-13
- Gemini CLI PR fixes sandbox: trap-wrapped commands broke per-command permission rules — MauroDruwel · 2026-10-13