AnalysisAI ModelsJuly 22, 2026
Relay-Bench: GPT-5.5 scores 43.3% on multi-domain reasoning benchmark
GPT-5.5 led with 43.3% on Relay-Bench, a new text-only benchmark evaluating multi-domain reasoning chains. The benchmark is designed to be unsaturated.
3 sources
More stories today
- SpaceXAI plans Imagine API 2.0 with unified image, video generation
- Cisco Outshift proposes 'Internet of Cognition' for multi-agent superintelligence
- Krea2 RetroAnime LoRA trained on 18,000 cel animation images
- Closing the data loop in AI-driven drug discovery
- Building the enterprise environment for agentic AI