PCFBench - Background Mapping: leaderboard

Metric: Exact-match accuracy (%; mapping 109 material names to the correct ecoinvent activity; agentic harness; from PCFBench, a product-carbon-footprint benchmark of 614 expert-labelled items; scored deterministically or by a temperature-0 judge; higher is better). Source: arxiv.org. Saturation forecast: Around December 2026. 9 models tracked.

Top models

#ModelScore
1Gemini 3.1 Pro (Preview)93
2Claude Opus 4.685
3DeepSeek V3.284
4Claude Sonnet 4.682
5GPT-5.580
6Claude Haiku 4.579
7GPT-5.4 Mini75
8Gemini 3 Flash68

Interactive version: theaggregate.ai/benchmark?slug=pcfbench-background-mapping · How It Works · Data refreshed daily, snapshot 2026-09-26.