PCFBench - Total Emissions (Direct): leaderboard
PCFBench (2026) is a diagnostic benchmark for product carbon footprint estimation, carving the workflow into independently scored tasks over 614 expert-labelled items. Share of products whose directly predicted total emissions fall within a factor of two of the declared total, in percent; higher is better.
Metric: Within-2x rate (%; share of 175 EPD products whose directly predicted total kgCO2e is within a factor of two of the declared total; 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 February 2027. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 77 | #91 |
| 2 | GPT-5.5 | 77 | #48 |
| 3 | Claude Opus 4.6 | 73 | #100 |
| 4 | Gemini 3 Flash | 73 | #159 |
| 5 | Claude Sonnet 4.6 | 63 | #158 |
| 6 | Claude Haiku 4.5 | 63 | #568 |
| 7 | DeepSeek V3.2 | 63 | #403 |
| 8 | GPT-5.4 Mini | 60 | #389 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=pcfbench-total-emissions-direct · How It Works · Data refreshed daily, snapshot 2026-11-12.