Alem (Medium): leaderboard
Alem (2026) tests open-ended multi-agent coordination: a team of LLM agents shares a Craftax-like survival and crafting world with procedurally generated tasks that need synchronised action, handovers and communication. This board is the medium difficulty, scored by the team total return as a percent of the maximum achievable reward; higher is better.
Metric: Total return (%): each agent's cumulative reward divided by the maximum achievable total reward (376) and averaged across agents on the Medium 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 July 2028. 13 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) (High) | 18.6 | #94 (Gemini 3.1 Pro (Preview)) |
| 2 | Gemma 4 31B (IT) (Thinking) | 10.1 | #342 (Gemma 4 31B (IT)) |
| 3 | Qwen 3.6 35B A3B | 7.7 | #462 |
| 4 | GPT-5.4 (High) | 7.4 | #123 (GPT-5.4) |
| 5 | Qwen 3.5 122B A10B | 6.9 | #331 |
| 6 | Qwen 3.6 27B | 6.6 | #341 |
| 7 | Qwen 3.5 27B | 5.1 | #359 |
| 8 | gemma-4-26B-A4B-it (Thinking) | 4.9 | |
| 9 | Qwen 3.5 9B | 3.6 | #803 |
| 10 | Qwen 3.5 35B A3B | 3.3 | #482 |
| 11 | Gemma 4 E4B (IT) (Thinking) | 2.7 | |
| 12 | Llama 3.3 70B Instruct | 1.9 | #1357 |
| 13 | Llama 3.1 8B Instruct | 0.8 | #4246 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=alem-medium · How It Works · Data refreshed daily, snapshot 2026-11-13.