Drex Router picks the right LLM in 57-87ms at 87.6% accuracy, 10x cheaper than rival
rohanpaul_ai · x · 2026-11-22
NaceAI launched Drex Router, a small model (12B, built on Gemma) that reads a prompt plus your model list and returns its pick in 57-87ms — fast enough to sit in front of nearly any LLM call.
On 1,400 general-knowledge exam questions it scored 87.6% accuracy at $0.10 per 1,000 questions, versus rival Jev Router's 86.4% at $1.09 — roughly a 10x cost gap.
Unlike routers that only return a model name, Drex outputs a probability for every candidate plus a short reason, so callers can tell a confident pick from a coin flip. A benchmark chart shows it leading same-size decision models across Knowledge, Language, Retrieval, Tools and Arts categories.
More from coding & agent
- Dev sends 7 PRs to open-source coding agent Shelley after switching to Gemini — swaroopch · 2026-11-22
- Databricks previews agent platform that lets you git-branch your entire agent — matei_zaharia · 2026-11-22
- Taste MCP reins in AI's tendency to bloat UI with unneeded text and elements — annetgriffin · 2026-11-22
- Daniel Lemire: why you should write your own pull-request messages — lemire · 2026-11-22
- Codex bot takes #1 on Victoris Arena leaderboard for the first time — tristanbob · 2026-11-22
- Artist opens artwork to AI agents via a keyless remote MCP server — ranajideb · 2026-11-22