# Hightouch tests show cheaper model routing can approach Opus accuracy

_Published Friday, November 6, 2026 at 4:19 PM EST · AI · Latest · Tier 2 — Notable_

![LLM Judges for Identity Resolution at 4x Lower Cost | Hightouch — Primary](https://cdn.sanity.io/images/pwmfmi47/production/9a5d7191d614c824c822ec90672adaf424699d3d-2400x1256.png?w=1200)

Hightouch found that routing uncertain Jev decisions to Claude Opus 5.5 achieved 98.8% accuracy at $0.88 per 1,000 identities in a synthetic test of whether customer records incorrectly combine different people. Opus alone achieved 99.0% at $3.86 per 1,000.

The system sent cases with Jev confidence scores of 30-90% to Opus for a second opinion. On a harder test, the same approach achieved 95.3% accuracy at $2.94 per 1,000, compared with Opus alone at 95.1% and $5.39. The tests used simulated records with added errors; costs and latencies reflect Hightouch's own setup.

## Sources

- [hightouch.com](https://hightouch.com/blog/llm-judges-identity-resolution)

---
Canonical: https://techandbusiness.org/newswire/CGuWyTobCL5v6HLHoT3co7
Published: 2026-11-06T21:19:26.282Z
Story chronology: 2026-11-04T14:44:00.000Z
Retrieved: 2026-11-06T22:48:33.356Z
Publisher: Tech & Business (techandbusiness.org)