MultiSessionCollab - MATH-Hard: leaderboard
Metric: Task success (%; accuracy of the simulated user's final draft answer, MATH-Hard problems, 100 simulated users x 20 sessions, agent without memory). Source: arxiv.org. Saturation forecast: Around December 2026. 4 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-OSS-20B | 41.45 |
| 2 | Qwen 2.5 7B Instruct | 26.5 |
| 3 | Llama 3.3 70B Instruct | 25.81 |
| 4 | Llama 3.1 8B Instruct | 11.5 |
Interactive version: theaggregate.ai/benchmark?slug=multisessioncollab-math-hard · How It Works · Data refreshed daily, snapshot 2026-09-26.