Vibe-Bench - BERT Retrieval - Preference Recall: leaderboard

Metric: Score (0-100). Source: arxiv.org. Saturation forecast: Around December 2026. 6 models tracked.

Top models

#ModelScore
1Llama 3.1 8B Instruct90.62
2Qwen 3 235B A22B 2507 Instruct90.62
3Llama 3.1 70B Instruct70.31
4Mistral Small 3.264.06
5Qwen 3 14B62.5
6Gemma 3 27B (IT)56.25

Interactive version: theaggregate.ai/benchmark?slug=vibe-bench-bert-retrieval-preference-recall · How It Works · Data refreshed daily, snapshot 2026-09-24.