MAC Bench - Team of Eight: leaderboard

Bespoke Labs' Multi Agent Coordination Bench: twelve four-hour tasks built so that each agent's work depends on its teammates' (air traffic control, a Mars sample-return mission, restoring a national grid after a blackout, an AI player for the strategy game 0 A.D.), scored 0-100 by a deterministic grader the agents cannot see. This board is a team of eight agents of one model with a shared message board, in the model's native harness; the companion boards run the same model alone and as eight agents that cannot message, which isolates what coordination itself adds.

Metric: Mean final score x100, team of eight agents with a shared message board. Source: benchmarks.bespokelabs.ai. Saturation forecast: Around August 2027. 3 models tracked. Updated 19 Nov 2026.

Top models

#ModelScoreOverall rank
1Claude Opus 5.553.38#9
2GPT-6.1 Sol40.58#15
3Gemini 4 Argon31.87#18

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

Interactive version: theaggregate.ai/benchmark?slug=mac-bench-team-of-eight · How It Works · Data refreshed daily, snapshot 2026-11-20.