SciImpact - Code: leaderboard

SciImpact pairs of code repositories that differ in GitHub stars, from their READMEs; the model sees two artifacts from the same field and picks the one with higher impact; pairwise accuracy in percent, chance 50, higher is better.

Metric: Pairwise prediction accuracy (%, times 100) on code repository pairs (GitHub stars, from READMEs), over SciImpact contrastive pairs (215,928 pairs over 19 fields, built from citation counts, best paper awards and Nobel prizes, patent citations, media attention, GitHub stars and Hugging Face downloads with impact thresholds): the model sees two same-field artifacts (title and abstract, README, dataset card or model card, truncated to 1,000 words) and names the one with higher future impact in a two-option forced choice parsed by exact match; the order is balanced so chance is 50%; higher is better. Source: arxiv.org. Saturation forecast: Around May 2028. 11 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1O4 Mini65.8#368
2Claude Haiku 4.562.6#583
3GPT-4.1 Mini60.3#789
4Qwen 2.5 14B57.7#1811
5Qwen 3 4B57.3#2348
6Qwen 2.5 7B56.3#2929
7Llama 3 8B54.7#5900
8Ministral 3 3B54.2#3504
9nemotron-3-nano-30B-a3B52.8#1946
10Llama 3.1 8B52.5#5196
11Llama 3.2 3B51.3#7877

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=sciimpact-code · How It Works · Data refreshed daily, snapshot 2026-11-21.