Alem (Hard) - Coordination: leaderboard

Metric: Coordination return (%): each agent's reward from coordination tasks divided by its maximum (159) and averaged across agents on the Hard difficulty, zero-shot homogeneous team (every agent driven by the same LLM) in alem, a JAX Craftax-like open-ended multi-agent world with procedurally generated coordination tasks (synchronous, all-agent and handover) and soft role specialisation; text interface with the rules, local observation and legal actions each step, scratchpad memory, broadcast communication and the last eight observations; reasoning enabled where available, proprietary models at high reasoning; mean over 20 seeds (10 for the API models); higher is better. Source: arxiv.org. Saturation forecast: Around 2030. 13 models tracked.

Top models

#ModelScore
1Gemini 3.1 Pro (Preview) (High)17.5
2gemma-4-26B-A4B-it (Thinking)9.5
3Gemma 4 31B (IT) (Thinking)8.8
4Qwen 3.6 35B A3B8.7
5Qwen 3.5 122B A10B8.3
6Qwen 3.5 27B8.1
7Qwen 3.6 27B7.2
8Qwen 3.5 35B A3B5.3
9Qwen 3.5 9B5
10Llama 3.3 70B Instruct4.2
11GPT-5.4 (High)4.2
12Gemma 4 E4B (IT) (Thinking)3.2
13Llama 3.1 8B Instruct0.3

Interactive version: theaggregate.ai/benchmark?slug=alem-hard-coordination · How It Works · Data refreshed daily, snapshot 2026-09-29.