HELM SeaHELM - LINDSEA Scalar Implicatures (id) — leaderboard

Metric: EM. Source: crfm.stanford.edu. 21 models tracked.

Top models

#ModelScore
1Llama 3.3 70B Instruct92.31
2Llama 3.1 70B Instruct91.35
3DeepSeek V387.5
4Claude 3.7 Sonnet (20250219)87.5
5Qwen 2.5 72B Instruct86.54
6Llama 4 Scout Instruct86.54
7Gemini 2.0 Flash Lite86.54
8Qwen 2 72B Instruct85.58
9Gemini 2.0 Flash80.77
10Llama 3.2 90B Vision Instruct74.04
11GPT-4o (2024-11-20)72.12
12Llama 4 Maverick Instruct FP848.08
13Llama 3.1 405B Instruct42.31
14GPT-4o Mini (2024-07-18)25.96
15Gemma 2 27B (IT)16.35

Interactive version: theaggregate.ai/benchmark?slug=helm-seahelm-lindsea-scalar-implicatures-id · How the rankings work · Data refreshed daily, snapshot 2026-07-22.