MCLM - MT-AIME2024: leaderboard

Each model solves the 30 AIME 2024 problems machine-translated by GPT-4o into 55 languages (1,650 problems per model), with temperature 0 and at most 3,276 output tokens, checked by a rule-based answer verifier; the score is the mean of the 55 per-language accuracies. It is the competition-level sibling of MCLM's MT-MATH100 subset and differs from English AIME 2024 boards by averaging over languages; the short output cap limits reasoning models.

Metric: Accuracy (%). Source: arxiv.org. Saturation forecast: Estimated already saturated. 5 models tracked. Updated 17 Nov 2026.

Top models

#ModelScoreOverall rank
1O3 Mini (2025-01-31)45.33#498
2GPT-4o Mini (2024-07-18)20.18#1494
3Qwen2.5-Math-7B-Instruct20.12#8400
4DeepSeek R1 Distill Qwen 1.5B17.21#9148
5Qwen2.5-Math-1.5B-Instruct16.36#9038

No result here: #1 Claude Fable 5.2, #2 GPT-6.2 Astra, #3 Gemini 4 Carbon.

Interactive version: theaggregate.ai/benchmark?slug=mclm-mt-aime2024 · How It Works · Data refreshed daily, snapshot 2026-11-17.