G-IdiomAlign - Multiple-Choice Equivalence: leaderboard

Metric: Accuracy (%; source idiom to its meaning-equivalent target-language idiom among four options, the other three typed distractors (literal translation, lexical cue, contextual association), greedy decoding; macro-average over all 30 directions). Source: arxiv.org. Saturation forecast: Around December 2026. 7 models tracked.

Top models

#ModelScore
1Gemini 2.5 Pro65.13
2DeepSeek V3.2 (Thinking)61.45
3DeepSeek V3.2 (Non-reasoning)52.7
4Claude Haiku 4.550.5
5Qwen 3 8B (Thinking)36.14
6Qwen 3 8B (Non-reasoning)28.98
7Ministral-8B-Instruct-241028.68

Interactive version: theaggregate.ai/benchmark?slug=g-idiomalign-multiple-choice-equivalence · How It Works · Data refreshed daily, snapshot 2026-09-26.