Wire · opportunities
Stanford-Oxford-MIT Benchmark Caps Top AI Teammate at 82.2
◆ Sectors
AI Frontier Models
◆ Geography
United StatesUnited Kingdom
◆ Source
◆ Verified
Fusion42 · 3 November 2026 · Fusion42 review
A new benchmark by Stanford, Oxford, and MIT shows top AI models score only 82.2 out of 100 as effective teammates, reaching the best team deal in just 14% of conflicts despite frequent inquiries about teammates' needs.
This Wire brief sits within Fusion42's coverage of AI Frontier Models.
◆ ◆ The Wire takeaway
You cannot rely on top AI models to negotiate or collaborate effectively under pressure yet. Focus on building AI that actively drives outcomes rather than just responding or conceding in conflicts.
◆ Coverage
1 source · 3 Nov 2026
◆ Related on Wire
◆ Topics
AI Frontier Modelsfrontier-modelsteamworkbenchmarkai-cooperationconflict-resolution