MultiSessionCollab - MATH-500: leaderboard
Metric: Task success (%; accuracy of the simulated user's final draft answer, MATH-500 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 | 67.6 |
| 2 | Llama 3.3 70B Instruct | 59.29 |
| 3 | Qwen 2.5 7B Instruct | 51.35 |
| 4 | Llama 3.1 8B Instruct | 44.55 |
Interactive version: theaggregate.ai/benchmark?slug=multisessioncollab-math-500 · How It Works · Data refreshed daily, snapshot 2026-09-26.